eSTEPS EHR Decision Support Linked to Fewer Coded Falls
Key Takeaways
- Among older US primary care patients with positive fall-risk screens, eSTEPS practices had modestly fewer ICD-10–coded falls than usual-care practices, a statistically significant difference in adjusted analyses.
- Coded fall-related injuries were also modestly and significantly lower at intervention practices in adjusted analyses.
- Exercise prescriptions reached only a minority of intervention patients, although most intervention clinicians used the decision support at least once.
- Mortality did not differ between intervention and usual-care groups.
Investigators in the eSTEPS cluster randomized trial matched and randomized primary care practices in pairs within Mass General Brigham, a US integrated health system. The 22 analyzed practices included 11,945 community-dwelling adults aged at least 65 who screened positive for fall risk at an index visit or during the preceding year. Usual care included routine fall-risk screening and the option of physical therapy referral. The prespecified primary outcome was the rate of patients with at least one EHR-derived International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10)–coded fall per 100 patient-years, not a count of every fall or repeat event. Coded fall-related injury and mortality were secondary outcomes.
Within Epic's EHR, a noninterruptive care-gap alert made eSTEPS clinical decision support (CDS) available after a positive fall-risk screen. Clinicians could use the associated order set to prescribe tailored, home-based strength and balance exercises and optionally refer patients to physical therapy; use remained discretionary. Handouts, after-visit and patient-portal materials, a website, and a mobile app offered ways to deliver exercise information, but their availability did not establish whether patients exercised. Control practices lacked the eSTEPS alert and order set.
Patients at intervention practices had a rate of 19.3 with at least one ICD-10–coded fall per 100 patient-years, versus 21.2 at usual-care practices. The rate ratio (RR) was 0.91 (95% confidence interval [CI], 0.88–0.95). A person with repeated falls did not contribute a separate event for each fall.
For the secondary injury outcome, the corresponding rates were 15.7 versus 16.7 patients with an ICD-10–coded fall-related injury per 100 patient-years. The RR was 0.94 (95% CI, 0.90–0.97). These diagnostic-code measures did not capture every fall or injury experienced outside the health record.
After adjustment for prespecified patient- and practice-level covariates, coded falls (P < .001) and fall-related injuries (P = .001) were significantly lower; unadjusted analyses did not show a treatment effect for these outcomes. Separately, an adjusted model estimated the probability of at least one fall over two years at 32% in the intervention group versus 34.5% in the control group. Those probabilities describe modeled risk, not the observed rates per 100 patient-years.
The authors reported a two-year number needed to treat (NNT) of 28.6 (95% CI, 19.0–42.9) to prevent at least one fall, not every fall an individual might experience; however, that figure does not reconcile with the reported modeled probabilities of 32% versus 34.5%. Exercise prescriptions were issued to 1,255 intervention patients (22.1%); most intervention clinicians used the CDS to deliver the program at least once. Safety monitoring identified no intervention-associated serious adverse events or unanticipated trial events.
Diagnostic codes could miss falls, and differences in reporting or documentation could affect which events appeared in the EHR. Falls or injuries reported during routine screening also differed significantly in favor of intervention practices. Exercise prescribing and clinician use of CDS varied by practice, while patients' exercise uptake and adherence were not measured. Physical therapy referral proportions were similar across groups, and the contribution of individual eSTEPS components or other clinician behaviors could not be isolated. The intervention was assessed within one integrated US health system using one EHR platform, leaving its performance elsewhere unresolved.
The investigators described the differences in coded falls and injuries as modest. They called for work to improve intervention uptake and workflow integration and to evaluate effectiveness across diverse practice settings and multiple EHR systems.
Clinician Questions
Which fall-risk screen responses made older adults eligible for eSTEPS in primary care?
Community-dwelling adults aged 65 or older qualified after answering yes to at least one question at an index primary care visit or a visit in the previous 12 months: whether they had a fall with injury in the past year, had two or more falls in the past year, or feared falling because of balance or walking problems. The responses were alternatives, not cumulative requirements.
Was hospitalization a prespecified outcome for eSTEPS fall prevention in older primary care patients?
Investigators assessed hospitalization among older primary care patients in the eSTEPS trial as an exploratory post hoc outcome, not as the prespecified primary coded-fall outcome or the secondary coded-injury and mortality outcomes. The adjusted hospitalization rate was slightly lower at intervention practices; this observation remains distinct from the prespecified findings.