Research & Validation
What makes the Hendrich Model unique?
With active and ongoing research, there is simply no other fall risk tool that has been validated in as many patients, at as many diverse sites of care, and with longer duration of study than the Hendrich II Fall Risk Model®. With dozens of independent replications studies across the care continuum, Hendrich is notably the most validated tool in the industry.
What makes the Hendrich program distinctive is the level of evidence from multiple studies, diverse sites, and over many years, the strength of evidence-based interventions provided for each validated and modifiable risk factor, and the comprehensive nature of the overall program.
What the research shows
The model was developed over a two-year study enrollment period with 355 patients who fell and 780 non-fall patients (sample size = 1,135). Each study participant was assessed concurrently by four registered nurses who used validated standardized instruments to test more than 600 variables. A clinical pharmacist categorized and evaluated all medications. Statistical measures used to generate the model included manual stepwise logistic regression, two-way interactions, 2 Log L chi-square statistic, and sensitivity and specificity. The validated model showed a sensitivity of 74.9% (falls correctly predicted) and a specificity of 73.9% (non-falls correctly predicted).
The research identified eight risk factors necessary to prevent falls. Interestingly, age and a history of falling were not independent risk factors for falling. Some older adults have conditions that put them at risk for falls, but age alone is not causal or predictive of falls. To accurately identify which older adults are more prone to fall, we must use weighted, evidence-based risk factors, many of which can be managed with targeted interventions.
2020 validation study
In 2020, the Hendrich team published the largest study of a fall risk model to date, with a study population of over 214,000 patients at nine sites of care and a study duration of three years. The nine hospitals included seven acute care sites with trauma populations and two critical access sites. Bed numbers at the included sites ranged from 25 to 474, with all levels of acuity and types of inpatient care, including behavioral health, skilled nursing, observation, and emergency departments. Study hospitals had patient populations representative of national diversity for race, gender, age, ethnicity, case-mix, length of stay, licensed bed size, academic and nonacademic settings, and urban and rural locations.
This study confirmed the Hendrich model’s accuracy in predicting falls. The Hendrich model demonstrated a sensitivity of 78.72%, specificity of 64.07%, and an AUC of 0.765 at risk score ≥ 5. Note that this study was conducted in hospitals with established fall prevention programs and fall reduction goals and a low overall fall rate of 0.29%. This means that the study likely underestimates the predictive accuracy of the Hendrich model.
Another important finding: The study found that a third of adult inpatients had multiple fall risk factors (n = 77,292), which clinical teams often do not manage during hospitalization. This finding points to a large opportunity for health care teams to reduce injurious falls, both in the hospital and post-discharge, through the active management of these risk factors.
A comparison of fall risk tools in the peer-reviewed literature
| Model Development | Hendrich | Hester Davis | Johns Hopkins | Morse |
|---|---|---|---|---|
|
Multisite study for tool development and testing |
|
|
|
|
|
Patient sample |
1082 falls |
44 falls |
261 falls |
247 falls |
|
Diversity of facility size and type |
Critical access through Level 1 trauma |
One 30-bed neurosciences unit in one hospital |
18 units in a tertiary hospital |
Large acute care hospital and rehabilitation hospital |
|
Selection of scale items |
Physical assessment of patients |
Literature review |
Literature review |
Physical assessment of patients |
|
Length of study |
2- and 3-year studies |
10 months |
14 months |
4-5 months |
|
Sensitivity (falls correctly predicted) |
78.72-90.56% based on cutoff score |
90.9% |
87.4% |
78% |
|
Specificity (non-falls correctly predicted) |
64.07% |
47.1% |
28.1% |
83% |
|
Validation |
2003 and 2020 |
2013 |
2018 |
1989 |