Stakeholder and Data Driven Fall Screen in a Program of All-Inclusive Care for the Elderly: Quality Improvement Initiative

Abstract

Background and Purpose:

The Program of All-Inclusive Care for the Elderly (PACE) delivers community-based long-term care services to low-income, nursing-home eligible adults. In the PACE population, one of the most common reasons for hospitalizations is falls. The purpose of this quality improvement study was to create a stakeholder driven process for developing a fall risk screen and evaluate how well this process discriminated injurious and non-injurious fallers.

Methods:

The quality improvement design was a prospective, longitudinal data collection for 5 PACE programs in Colorado. Physical Therapists collected the Short Physical Performance Battery (SPPB) on participants at least annually. The Kotter practice change framework guided the processes for practice and organizational change in developing and implementing a fall screen.

Results and Discussion:

An iterative, stakeholder and data-driven process allowed our team of researchers and a PACE program to establish a fall risk screen to stratify PACE participants. We provided feedback to PACE staff regarding screening rates and results on discrimination of faller status to promote continued uptake of screening and discussion regarding next steps.

Rehabilitation therapists screened 66% of the PACE population, and participants were stratified into high risk (1–7 points) or low risk (8–12 points) on the SPPB. Participants with low SPPB scores had 79% greater risk of a fall (RR: 1.8, 95% CI: 1.5–2.1) and 86% greater risk of an injurious fall (RR: 1.9, 95% CI: 1.4–2.4), compared to those with high SPPB scores.

Conclusions:

Our study describes a collaboration to address fall rates in a PACE population. PACE clinicians can use the identified cut-offs to stratify PACE populations at risk for falls and allocate scarce rehabilitation resources efficiently to intervene on participants at highest risk, while using less resource-intensive interventions for those at lower risk.

INTRODUCTION

The Centers for Medicare & Medicaid Service’s (CMS) Program of All-Inclusive Care for the Elderly (PACE) delivers community-based long-term care services to low-income adults who are eligible for nursing home admission.^1,2 Under PACE, interdisciplinary services—including physical and occupational therapy--are provided through a capitated payment model, with most participants in PACE models being dual eligible for Medicare and Medicaid. In this vulnerable population, falls are a major outcome of interest. Falls can initiate a negative sequela of events including hospitalization, loss of independent mobility, costly institutionalization in a nursing home, or death.^3–7 In fact, falls cost 50 billion USD in 2015, taxing an already burdened healthcare system.⁴ In addition, a fall-related hospitalization in long-term care and PACE settings⁸ is considered a potentially preventable hospitalization, which impacts PACE facility quality metrics. Therefore, developing strategies for identifying and intervening on modifiable fall risk factors is of critical concern in the PACE population.^9–12 However, PACE programs lack standardized guidelines for managing fall risk and have historically relied on individual clinical judgement and a myriad of fall risk screening tools. Any fall risk screening tools may not be specific enough for the nursing-home eligible PACE population, resulting in over identification of those at imminent risk for falls and poor tailoring and targeting of resources to address those participants most likely to benefit from interventions.

Impairments in physical function are actionable and modifiable risk factors for falls. Muscle weakness, slow gait speed, and impaired static balance are all sensitive risk factors for falls that are amendable to interventions.¹³ However, these factors are not typically assessed in a standardized fashion in PACE participants, which represents a missed opportunity to prevent falls in this vulnerable population. Therefore, the purposes of our quality improvement study were to 1) develop an iterative, stakeholder engaged, data-driven process for developing and implementing a fall risk screening protocol guided by the Kotter practice change framework,^14,15 and 2) evaluate whether the developed fall risk screen successfully discriminated different levels of risk for falls and injurious falls in the PACE population.

METHODS

Sample Characteristics and Data Collection

InnovAge is a PACE provider with 5 sites in Colorado, as well as additional sites in Virginia, California, and Pennsylvania. We collected data from April 2015, through August 2017, in the 5 Colorado InnovAge PACE sites that provided long-term care services to nearly 3000 eligible beneficiaries. Ethical approval was obtained from the Colorado Multiple Institutional Review Board (#15-0712). To guide and describe the processes for practice and organizational change, we used the Kotter practice change framework. Briefly, the Kotter framework includes 8 steps for practice change: create a sense of urgency, build a guiding coalition, create a vision, involve everyone in the vision and plan, reduce barriers, focus on short-term wins, sustain momentum, and add stability.^14,15 To screen for falls risk, rehabilitation therapists administered the Short Physical Performance Battery (SPPB) on all PACE participants at standard annual, bi-annual physical assessments, or within 30 days of a reported fall or hospitalization. The SPPB consists of 3 sections: static balance test, gait speed, and a timed 5-time sit-to-stand; each section is scored 0–4 on an ordinal scale where higher scores indicate better function.¹⁶

Data extraction was conducted by the InnovAge analyst team approximately quarterly from the electronic medical records and billing records. Demographic data were extracted from billing and administrative records and SPPB score data were pulled from an annual physical assessment documented in the medical record by the rehabilitation staff. Stakeholders also identified the importance of knowing a participant’s medical complexity and dementia as these factors were perceived to play an important role in clinical decision-making regarding fall risk treatment and prevention. Hierarchical condition category (HCC) coding and dementia diagnosis were extracted from billing and administrative records. HCC codes measure multimorbidity that represents burden of chronic disease and is used in the PACE program to help risk-adjust quality and cost metrics.¹⁷ Fall number and nature (i.e., injurious versus non-injurious) information was extracted from detailed fall incident reports filled out by clinical staff. Injurious falls were categorized by documentation in the incident report as any fall-related harm (i.e., minor, moderate, or severe) resulting in a need for hospitalization, emergency department visit, or medical attention (e.g., evaluation or office-based medical intervention) from a physician or nurse.

Data Analysis

Descriptive statistics for all participants were calculated for basic demographic data using median (interquartile range) for continuous data and frequency counts for categorical data. For each participant, the total number of falls were summed over the total time at risk, defined as the number of days between the first SPPB score and either the end of data collection (August 30, 2017), date of death, or date of disenrollment from the PACE program. We calculated fall rates for participants across the disability strata using Poisson regression models with an offset for the natural logarithm of total time at risk, defined as the number of days between the first SPPB score and either the end of data collection, date of death, or date of disenrollment from the PACE program. Relative risks and 95% confidence intervals were calculated to preliminarily assess whether the screening strategy successfully discriminated fall rates. The rate of falls is expressed in falls per 10,000 person days. All data analysis was completed with SAS version 9.4 (SAS, Cary, NC).

RESULTS

Implementation Process Using the Kotter Practice Change Framework

Create a sense of urgency: We collaborated with PACE rehabilitation clinicians and administrative stakeholders to determine fall-related quality improvement goals for rehabilitation and assess contemporary practice patterns related to fall risk screening. Collaborative discussion between stakeholders revealed falls were an overwhelming prevalent and concerning issue for PACE participants given the negative sequela stemming from a fall. Furthermore, the Center for Medicare and Medicaid Services’ capitated payment model for PACE creates an incentive to prevent costly hospitalizations or institutionalizations following a fall. Stakeholders agreed that a common fall risk screening tool was needed to facilitate better communication within rehabilitation and across to other disciplines (e.g., primary care) for improved safety of PACE participants. Of importance to stakeholders was the ability to risk-stratify participants to intervene efficiently and proactively on fall risk to prevent or delay time to a fall.

Build a guiding coalition: Our guiding coalition consisted of clinical researchers (3 PTs), administrative leaders (Director of Rehabilitation and Medical Director), and clinical champions (5 Physical Therapy Leads). We defined clinical champions as the PTs who took the lead on setting up the testing environment at each site, coordinated testing with other therapists, trained new therapists on the SPPB screening, and helped compile screening statistics at each site.

To address the issue of falls in PACE participants, we participated in multiple, iterative meetings to determine the best measures to assess physical function in PACE participants and strategies to embed standardized assessment of function into clinic flow. The research team presented options for functional measures including the pros and cons of each to allow stakeholders to make an informed decision regarding which measure would be administered at standard semi-annual or annual physical assessments. The SPPB was selected as the outcome of choice as it is a well-accepted composite measure of lower extremity function, is easy to administer in the home or clinic setting, and is a strong predictor of disability, institutionalization, fall related trauma, and morbidity in older adults.^8,9,16,18,19 The SPPB is reliable with intra class correlation coefficients (ICC) >0.88 and demonstrates good sensitivity to change.²⁰ Stakeholders met at least monthly for the first year, and clinical stakeholders provided valuable input on the initial implementation and ongoing evaluation of compliance with administering the SPPB at physical assessments to ensure all participants were screened.

Form a strategic vision and initiative: We collectively decided to move forward with a prospective, longitudinal data collection of objective measures of physical function on all participants to characterize the PACE population, inform clinical practices, and provide insight to future innovations to improve care. We sought to create a system of rapid data collection and statistical programming that was sustainable for long-term evaluation of trends and patterns in the InnovAge PACE population, with a vision to prolong participant time in the community.

Involve everyone in the plan: The research team provided a 1-hour training on SPPB administration to all PTs and physical therapist assistants across the 5 InnovAge sites to ensure reliable data collection and clinical interpretation. Following training on the SPPB, research team members met with each of the five site teams to verify reliable data collection of the SPPB by working with each site to set-up a common testing environment (i.e., standard chair, standard walking path), directly observing SPPB administration of all therapists at least once, and rapidly addressing any questions regarding the test. Questions from the therapy teams were compiled, answered, and shared with all therapists to ensure consistency in measurement prior to the commencement of data collection. Onsite meetings occurred monthly for the first year of data collection with observations and audits of SPPB administration for data accuracy and completeness. SPPB training was offered twice annually to all InnovAge therapists.

Reducing barriers: To reduce barriers related to data collection and reliability, research team members were available for questions during scheduled site visits or anytime by email or phone. After the initial data collection, all stakeholders met again to discuss appropriate cut points based on clinical experience and research⁸ to identify risk for falls and fall-related trauma in this long-term care population, where data is lacking for falls risk stratification and specification. Through iterative discussion, we agreed on a cut point (SPPB 8–12 and 0–7) and discussed concerns about non-ambulatory participants (SPPB=0). We presented the initial data and fall risk screen to all the rehabilitation teams at the five InnovAge sites during a one-hour meeting. Therapists identified the need to remove non-ambulatory participants from the fall risk screen (e.g., wheelchair-bound at baseline) because their risk for falls was lower and rehabilitation resources were taxed.

Focus on short-term wins: The research team met with all sites individually at least monthly during the first year of data collection to provide feedback on the reliability and completeness of data collection. We focused on how well teams were doing by reporting the volume of data received and provided constructive feedback and facilitated goal setting to rectify gaps in the data collection. Additionally, high administrative engagement fostered electronic medical record (EMR) integration within the first year, which drastically improved the quality and consistency of data collection.

Sustain momentum: The research team initially collected the paper SPPB forms at least monthly from all sites. During this time, InnovAge integrated the SPPB test into its electronic medical record (EMR) system, and PTs began documenting SPPBs electronically for extraction. Integration into the EMR, specifically in the physical assessment template, served as a prompt for PTs to complete the SPPB and better served their workflow as they completed other required items on the physical assessment. We presented the screening rates and results regarding discrimination of faller status to the clinical stakeholders to promote continued uptake of screening. To date, InnovAge PACE continues to collect SPPB data during physical assessments and has disseminated the process to sites outside of Colorado. Furthermore, researchers have optimized the statistical programming to allow for future, rapid analysis to update and refine the fall screen with stakeholder input.

Using data to drive stability & sustainability: We used the preliminary data from our quality improvement implementation as part of an iterative feedback loop to clinicians to celebrate success of their program and encourage sustainability.

From the initial implementation of the fall risk screening, we collected data across 5 InnovAge PACE sites in Colorado. A total of 1772 participants were screened from an eligible pool of 2694 participants with complete demographic data at the time of study start (a screening rate of 66%). Table 1 outlines the InnovAge PACE participant characteristics by SPPB category determined as high (8–12 points) or low (1–7 points). There were 162 participants excluded from fall risk screen consideration because they scored a 0 on the SPPB. Overall, participants experienced 1446 falls over the follow-up period (421 in the high SPPB group and 1025 in the low SPPB group). Mean age of participants at first SPPB assessment was younger at 72 years for the high SPPB category and older at 77 years for the group with the lower, at-risk SPPB score. A lower proportion of participants in the high SPPB category were female compared with the low SPPB group. The average number of hierarchical condition category codes in the high SPPB group was lower at 5 (range 3–7) compared to 6 (range 4–9) in the low SPPB group. Table 2 depicts the falls incidence and relative risk for ambulatory InnovAge participants by SPPB high or low category. Incidence of falls over 10,000 person days was higher in the low SPPB (1–7 points) group for both injurious and non-injurious falls. Participants in the low SPPB category had 79% greater risk of experiencing a fall (Risk ratio [RR] 1.8, 95% CI 1.5–2.1) and 86% greater risk of an injurious fall (RR 1.9, 95% CI 1.4–2.4), compared to participants in the high SPPB category. From the data and stakeholder input, Figure 1 was created as a fall risk screen for InnovAge participants.

Table 1.

PACE participant descriptive characteristics by SPPB^* category.

Variable	High SPPB 8–12 (N=567)	Low SPPB 1–7 (N=1043)
Age (years), median (range)	72 (66–79)	77 (69–83)
Sex (female), N (%)	64.3% (365)	73.1% (763)
Hierarchal Condition Codes, median (range)	5 (3–7)	6 (4–9)
Dementia, N (%)	29.6% (168)	39.2% (409)

^*SPPB: Short Physical Performance Battery

Table 2.

Falls incidence and relative risk for ambulatory PACE participants by SPPB category.

Variable	SPPB 8–12 (N=567)	SPPB 1–7 (N=1043)	Relative Risk Low vs High
Falls per 10,000 person-days (95% CI)	14.7 (12.7–17.1)	26.3 (22.1–31.3)	1.8 (1.5–2.1)
Injurious falls per 10,000 person-days (95% CI)	1.3 (1.1–1.7)	2.5 (1.9–3.2)	1.9 (1.4–2.4)

^*PACE: Program of All-Inclusive Care for the Elderly

^†SPPB: Short Physical Performance Battery

Figure 1.

Fall-screen assessment based on PACE participant data and stakeholder input. Participants can enter the risk assessment 1) during their annual/ biannual screening by Physical Therapists or 2) if they have 30-day history of a recent fall or hospital admission. Participants are screened using the Short Physical Performance Battery (SPPB) and placed into risk categories based on SPPB score. An SPPB score equal to zero or ≥8 indicates low risk for falls. SPPB scores between 1 and 7 points indicate high risk for falls and the need to intervene.

DISCUSSION

The initial quality improvement study to create a data- and stakeholder-driven falls screening tool for PACE participants led to an established and collaborative partnership between researchers and InnovAge. The short-term success of the screening tool development prompted renewed enthusiasm in challenging practice norms and creating a culture of innovation. In 2017, stakeholders convened again to discuss and strategize an approach to intervene on high-risk participants identified by the screening tool using progressive rehabilitation (manuscript in preparation). We continue to work together to address falls in the PACE population by combining and leveraging our unique skills, resources, knowledge, and experiences to achieve our collective vision to prolong community living for PACE participants.

This publication provides a successful example of engaging a large healthcare system in addressing a time-sensitive and clinically relevant problem: falls in PACE participants. An iterative, stakeholder and data-driven process allowed our research team and clinical collaborators to create a fall risk screen that can be used clinically to proactively identify participants at high risk of falling and intervene quickly. Using practice to inform research and data to inform practice is a continuous improvement cycle that creates a platform and culture for a learning health care system.^21,22 Establishing such an infrastructure and collaboration has the potential to increase the impact and timeliness of research on clinical practice and, importantly, provide rapid methods to address patient population needs and maximize outcomes. A key lesson learned is that embedding the SPPB administration in routine care through EMR integration is critical to ensuring all participants are screened for fall risk. We continue to work with stakeholders to insert the screening tool and interpretation into the EMR for rapid interpretation and documentation of fall risk.

Our secondary purpose was to identify relative risk for injurious and non-injurious falls based on longitudinal, objective measures of physical function. We found that ambulatory InnovAge participants who score between 1–7 points on the SPPB are greater risk for injurious and non-injurious falls compared to non-ambulatory participants and those who score ≥8 points. A SPPB score of ≤6 points in non-PACE, community dwelling older adults indicates risk for falls.²³ Our study demonstrated a cut-off of 7 points for an ambulatory PACE population, which may reflect the frailty, medical, psychosocial, and socioeconomic complexity of this population. Characteristics of PACE participants typically include eligibility for Medicaid services, older age (>55 years), female (~70%), limited education (~8 years), high number of chronic conditions (~7), and significant dependencies in activities of daily living (~4–7).^24,25 Taken together, these characteristics highlight a vulnerable population with minimal reserves to return to prior levels of function, level of assist, and community living status from the sequela associated with a the fall.

The causes of falls are multi-factorial; however, physical function is a significant, independent predictor of falls in PACE population²⁶ and, importantly, is a modifiable risk factor. The combination of movement systems evaluation and longitudinal monitoring of physical function offers rehabilitation the opportunity to contribute to the interdisciplinary InnovAge PACE team by risk-stratifying participants for falls and using information gathered in evaluation to recommend rehabilitation intervention (i.e., falls recovery training, strengthening and balance training, home assessment) or referral to other disciplines (e.g., pharmacy for medicine reconciliation). The fall risk screen provides clinicians with data-driven information to inform clinical decisions such as whom to target and when to intervene. Ideally, the fall risk screen stratifies InnovAge participants and allows for proactive intervention that can be tailored based on individual participants’ presentation, needs, and preferences.

This study has limitations to consider. First, data were collected on participants living near InnovAge sites in Colorado, which may not be generalizable to other InnovAge programs located in different geographic regions or other PACE programs. In addition, use of facility fallreporting data may undercount the total volume of falls, especially those that did not result in injury. This suggests our estimates of fall rates within PACE may be conservative.

CONCLUSIONS

This study described how researchers and clinical partners collaborated to solve a time-sensitive, highly relevant, and costly clinical problem in InnovAge PACE participants: falls. The combination of standardized, longitudinal data collection on participants’ physical function and clinician-level input allowed us to iteratively develop a fall risk screen with PACE population-relevant cut-offs for fall risk. Current quality improvement initiatives from our collaboration are exploring rehabilitation approaches to intervening on participants stratified as high risk by the fall screen. Future research will explore the identification and effectiveness of different clinical pathways to address the multi-factorial and diverse nature of falls for individual PACE participants. A targeted, data-driven approach is likely to better meet the needs of the patient, prevent or delay onset of adverse events, and overall be more cost-efficient for the healthcare system.

Clinical Implications.

• Falls risk cut-offs on measures of function for older adult populations may not be sensitive enough for a community-based, nursing home eligible population

• Assessment of physical function can help clinicians risk-stratify individuals and more efficiently intervene

• A stakeholder and data-driven approach to falls prevention may better meet the needs of the patient, provider, and system; future research is needed to identify specific clinical pathways for the diverse and multi-factorial nature of falls