Some But Not Too Much: Multiparticipant Therapy and Positive Patient Outcomes in Skilled Nursing Facilities

Abstract

Background and Purpose:

Physical and occupational therapy practices in Skilled Nursing Facilities (SNFs) were greatly impacted by the 2019 Medicare Patient Driven Payment Model (PDPM). Under PDPM, the practice of multiparticipant therapy – treating more than one patient per therapy provider per session – increased in SNFs, but it is unknown how substituting multiparticipant therapy for individualized therapy may impact patient outcomes. This cross-sectional study establishes baseline relationships between multiparticipant therapy and patient outcomes using pre-PDPM data.

Methods:

We used Minimum Data Set assessments from all short-term Medicare fee-for-service SNF stays in 2018. Using generalized mixed effects logistic regression adjusted for therapy volume and patient factors, we examined associations between the proportion of minutes of physical and occupational therapy that were received as multiparticipant sessions during the SNF stay and two outcomes: community discharge and functional improvement. Multiparticipant therapy minutes as a proportion of total therapy time were categorized as None, Low (below the median of 5%), Medium (median to <25%), and High (25% or more) to reflect the 25% multiparticipant therapy limit required by PDPM.

Results and Discussion:

We included 901,544 patients with complete data for functional improvement and 912,996 for the discharge outcome. Compared to patients receiving no multiparticipant therapy, adjusted models found small positive associations between low and medium multiparticipant therapy levels and outcomes. Patients receiving low levels of multiparticipant therapy had 14% higher odds of improving in function (95% CI 1.09–1.19) and 10% higher odds of community discharge (95% CI 1.05–1.15). Patients receiving medium levels of multiparticipant therapy had 18% higher odds of functional improvement (95% CI 1.13–1.24) and 44% higher odds of community discharge (95% CI 1.34–1.55). However, associations disappeared with high levels of multiparticipant therapy.

Conclusions:

Prior to PDPM, providing up to 25% multiparticipant therapy was an efficient strategy for SNFs that may have also benefitted patients. As positive associations disappeared with high levels (≥25%) of multiparticipant therapy, it may be best to continue delivering the majority of therapy in SNFs as individualized treatment.

INTRODUCTION

Over 1.5 million Medicare beneficiaries received post-acute short-term skilled nursing and rehabilitation services in Skilled Nursing Facilities (SNFs) in the United States in 2019.¹ Patients are discharged from hospitals to SNFs with a wide variety of medical and post-surgical diagnoses and are diverse in their specific post-acute care needs. The most common conditions for short-stay patients in SNFs in 2019 ranged from joint replacement and hip and femur surgeries to septicemia, heart failure, and kidney infections.¹ Despite heterogenous medical needs, the objectives of physical therapy (PT) and occupational therapy (OT) in SNFs are similar across patients admitted for short-stay rehabilitation, with common goals of improving patient mobility and self-care function and to facilitate discharge to a less restrictive care setting.²

Participation in rehabilitation during SNF stays is very common, and over 95% of Medicare beneficiaries received therapy services during post-acute SNF stays in 2018.³ However, rehabilitation in SNFs came under significant scrutiny by the Centers for Medicare & Medicaid Services (CMS) under the previous reimbursement system for SNFs, called the Prospective Payment System (PPS), which was in place between 1998 and 2019.⁴ The PPS incentivized higher intensity therapy provision by coupling higher payments to higher volumes of therapy, and CMS attributed rising costs of SNF care in large part to increased therapy utilization.^4,5 Under the PPS, Medicare SNF costs grew by $1 billion between 2010–2015,⁵ totaling over $28.5 billion by 2018.⁶ Concurrently, the share of patients receiving high levels of therapy rose from 27 percent to 84 percent between 2002–2018, with minimal changes in average patient functional dependence that would signal increased clinical need for therapy services.³ Additionally, equitable access to SNF services was a concern under PPS, as facilities were incentivized to admit patients who could tolerate high levels of rehabilitation and avoid patients with high medical complexity and nursing needs.⁵

Motivated by these PPS shortcomings, CMS implemented sweeping reforms to the SNF reimbursement system on October 1, 2019, replacing the PPS with the Patient Driven Payment Model (PDPM), which aims to decrease clinically unnecessary therapy in SNFs and re-align payment with patient need.⁷ To achieve these aims, PDPM determines reimbursement based on patient diagnosis and functional status and removes therapy minutes as the primary determinant of payment for Medicare fee-for-service beneficiaries, thereby disincentivizing higher rates of therapy provision.⁷

Consistent with previous SNF industry shifts in staffing after PPS implementation, it is anticipated that many SNFs will respond to changing incentives under PDPM by reducing therapy costs.^7–9 One cost-saving strategy is to replace individual therapy with multiparticipant therapy, defined as either group therapy (two to six patients per therapy provider doing similar activities) or concurrent therapy (two patients per therapist doing different activities) in contrast with individualized therapy (one patient per therapist per session).¹⁰ Multiparticipant therapy allows SNFs to provide therapy sessions to the same number of patients with fewer therapy staff and has thus been considered as a way to reduce labor costs.^11,12 Multiparticipant therapy was not incentivized under the PPS and made up less than 1% of total therapy minutes prior to PDPM.¹² The literature has not fully described how clinical decisions are made in terms of including certain patients in individual versus multiparticipant therapy sessions. However, prior to PDPM, there is evidence that SNF organizational factors were more strongly associated with multiparticipant therapy provision than patient clinical characteristics, thus these decisions are likely driven by a combination of clinician decisions and organizational policies.¹³

As anticipated, the SNF industry increased multiparticipant therapy under PDPM, both in terms of overall minutes of multiparticipant therapy provided as well as the proportion of patients receiving multiparticipant therapy. CMS reported that the percentage of patient stays that included concurrent therapy rose from 1% to 32% and the percentage receiving group therapy rose from 1% to 29% immediately after PDPM implementation.¹⁴ In a post-PDPM survey, 76% of SNF therapists and administrators reported increases in group therapy, and over two-thirds of respondents indicated that SNF employers either encouraged or mandated changes in multiparticipant therapy.¹⁵ While patient isolation measures necessary during the COVID-19 pandemic reversed the early increase in multiparticipant therapy, further increases in multiparticipant therapy provision in SNFs are likely after successful vaccination efforts.^1,14,16

Citing preference for individualized therapy as the dominant modality of therapy provision in SNFs and responding to stakeholder concerns about excessive multiparticipant therapy provision, CMS limited multiparticipant therapy to 25% of total therapy time under PDPM, however no financial penalties are incurred by SNFs who exceed the 25% limit.^7,17 While there is some evidence on the efficacy of multiparticipant therapy compared to individualized therapy, previous studies focused on specific patient populations (e.g., brain injury, back pain) in other healthcare settings, which are not generalizable to the heterogenous SNF population.^18–22 Only one study has examined relationships between multiparticipant therapy and rehabilitation outcomes in SNFs, finding no significant association between multiparticipant therapy provision and patient improvement in gait speed or balance.¹¹ However, this small study included patients from only one facility and used clinical outcome measures that CMS does not monitor as part of its quality initiatives. Another study found that low-quality SNFs were more likely to provide ≥25% multiparticipant therapy prior to PDPM, however the Five-Star quality measure used in this study provides a broad facility-level summary of overall nursing and rehabilitation quality in SNFs and may not accurately reflect relationships between multiparticipant therapy and patient-specific outcomes.^13,23 Thus, it remains unclear how national shifts in therapy practice away from individualized treatment towards multiparticipant therapy will impact patients, especially if increases in multiparticipant therapy are driven by organizational mandates rather than individual clinician decision-making. As post-PDPM patient-level data on therapy provision in SNFs are only available at two-year lags, the aim of this study was to establish pre-PDPM baseline relationships between multiparticipant therapy and patient outcomes in order to facilitate comparisons and provide insight into how patients could be affected as multiparticipant therapy increases in response to changing reimbursement incentives.

METHODS

Data and Study Population

This cross-sectional analysis used patient data from the calendar year 2018 Minimum Data Set (MDS) 3.0 and Master Beneficiary Summary File (MBSF), which include data on all SNF residents in the United States, specifically patient demographic, clinical, and functional data, as well as service and therapy use during SNF admission.²⁴ To avoid inducing correlation by using multiple stays per SNF resident, we included the first complete Medicare Part A fee-for-service SNF stay per patient in 2018. We excluded patients missing from the MBSF, long-stay patients with a length of stay over 100 days, and patients who were admitted and discharged on the same day.^25–27 Consistent with previous research on short-stay SNF patients, patients on hospice care or who were comatose at admission, patients who died during their SNF stay, and patients who received no therapy throughout the SNF stay were excluded.^25,28 For the functional improvement outcome, we then removed patients who did not have two or more assessments during their stay, as we could not calculate changes in function. For the community discharge outcome, we excluded patients with discharge location missing from their discharge MDS assessment.

Outcomes

The two dichotomous outcomes in this study are patient functional improvement and successful discharge to a community setting. Community discharge location is indicated on MDS assessments, and can include a private home or apartment, assisted living, or group home setting, in contrast with an acute care hospital, hospice, or another rehabilitation facility. This MDS indicator for community discharge is used by CMS and frequently used in studies examining post-acute care outcomes.^26,27,29,30 Consistent with CMS methods used in Nursing Home Compare and Quality Reporting Program initiatives, for this study, functional improvement occurred if a patient improved by at least one point on one or more of the six-point scales for transfers, ambulation, or locomotion performance between admission and discharge MDS assessments.³¹

Independent Variables

The independent variable of interest is multiparticipant therapy minutes as a proportion of total therapy minutes during the SNF stay. We added minutes of concurrent and group PT and OT across all MDS assessments and divided by total minutes of PT and OT across assessments completed throughout the SNF stay. As the overall proportion of multiparticipant therapy was anticipated to be small, we decided a priori to categorize multiparticipant therapy provision into levels that were clinically and policy relevant. We used the median proportion of multiparticipant therapy to distinguish between low and medium multiparticipant therapy and added an additional category for patients receiving at or above the CMS limit of 25% multiparticipant therapy. The median proportion among patients who received some multiparticipant therapy was 4.9%, so our final categories of multiparticipant therapy were: None (0%), Low (>0 to <5%), Medium (5 to <25%), and High (25% or more).

Covariates

Individual-level covariates were obtained from MDS admission assessments and the MBSF. Patient demographic covariates included age, sex, dual Medicare-Medicaid eligibility, marital status, need for an interpreter, and disability or end stage renal disease as the reason for Medicare entitlement. SNF stay factors included length of stay (LOS) in days, therapy intensity reflected by OT and PT minutes received per day of therapy, and indicators for specific treatments received during the SNF stay, such as intravenous medication and hemodialysis. Patient functional covariates coded in the MDS included the activities of daily living (ADL) scale score at admission, which reflects patient performance on seven functional tasks.³² The ADL scale ranges from 0 to 28 with higher scores indicating more severe functional impairment. We included covariates for use of an assistive device (either a wheelchair, walker, cane, or prosthetic), communication or vision impairment, falls in the last 6 months, and whether patients reported pain on a daily basis during admission assessments. We included scores on the Brief Interview for Mental Status (BIMS), with scores of 13 or higher classified as no cognitive impairment, scores of 8–12 classified as moderate impairment, and scores below 8 as severe cognitive impairment.^25,33 Other covariates include dichotomous indicators from admission assessments for behaviors such as wandering, rejection of care, and psychosis such as hallucinations or physical or verbal behavioral symptoms directed towards others.^24,29 Finally, we included a wide range of active diagnoses coded on the MDS, ranging from stroke and heart failure to hip fracture or Alzheimer’s dementia (Table 1).²⁴

Table 1.

Descriptive Statistics for Complete Short-Stay Skilled Nursing Facility Patient Stays in 2018

	Functional Improvement Outcome (N=901,544)	Community Discharge Outcome (N=912,996)
Multiparticipant Therapy	Mean (SD) or N (%)	Mean (SD) or N (%)
Proportion Multiparticipant PT + OT*	0.23 (1.9)	0.22 (1.9)
Proportion Multiparticipant PT	0.26 (2.3)	0.25 (2.3)
Proportion Multiparticipant OT	0.20 (2.3)	0.20 (2.3)
Multiparticipant Therapy Categories
None (0%)	873,421 (96.9%)	886,455 (97.1%)
Low (>0 to <5%)	14,510 (1.6%)	13,202 (1.4%)
Medium (5 to <25%)	12,494 (1.4%)	12,208 (1.3%)
High (25% or more)	1,119 (0.1%)	1,131 (0.1%)
Outcomes
Improved in Function	519,779 (65.0%)	-
Successful Community Discharge	-	687,573 (75.3%)
Demographics
Age (years)	80.6 (8.5)	80.3 (8.4)
Female	557,238 (61.8%)	556,253 (60.9%)
Disability or ESRD as reason for Medicare Entitlement	181 (0.02%)	203 (0.02%)
Dual Eligibility	164,158 (14.8%)	135,360 (14.8%)
Married	315,509 (35.0%)	339,124 (37.1%)
Needs Interpreter	23,048 (2.6%)	21,902 (2.4%)
SNF Stay Characteristics
Length of Stay (days)	27.8 (19.6)	22.8 (17.5)
PT Minutes per Stay	733.4 (394.3)	674.0 (379.0)
OT Minutes per Stay	690.9 (379.8)	631.8 (363.7)
PT Minutes per Day of Therapy	56.7 (11.1)	56.8 (11.4)
OT Minutes per Day of Therapy	55.4 (10.7)	55.4 (11.0)
Cognitive and Physical Function
ADL Scale Score on Admit (0–28)	16.7 (4.1)	16.7 (4.2)
Falls in the last 6 months	423,101 (46.9%)	408,100 (44.7%)
Daily pain	13,253 (1.5%)	13,071 (1.4%)
Use of an assistive device	899,925 (96.4%)	878,212 (96.2%)
Cognitive Impairment
None (BIMS 13+)	579,559 (66.8%)	591,548 (70.2%)
Moderate (BIMS 8–12)	175,962 (20.3%)	160,602 (19.0%)
Severe (BIMS 0–7)	112,384 (12.9%)	91,020 (10.8%)
Vision Impairment	133,209 (14.8%)	129,878 (14.2%)
Communication Impairment	47,601 (5.3%)	41,446 (4.5%)
Wandering	15,320 (1.7%)	12,592 (1.4%)
Rejected Care	38,756 (4.3%)	36,234 (3.7%)
Active Diagnoses
Heart Failure	207,029 (23.0%)	212,768 (23.3%)
Stroke	72,859 (8.1%)	68,297 (7.5%)
Hip Fracture	83,002 (9.2%)	80,296 (8.8%)
Other Fracture	111,131 (12.3%)	107,637 (11.8%)
Psychosis or Behavioral Symptoms	39,639 (4.4%)	35,400 (3.9%)
Wound Infection	6,884 (0.8%)	7,101 (0.8%)
Paraplegia	1,973 (0.2%)	1,860 (0.2%)
Multiple Sclerosis	4,024 (0.4%)	3,909 (0.4%)
Foot Infection	8,744 (1.0%)	8.858 (1.0%)
Anemia	241,815 (26.8%)	245,107 (26.8%)
Pneumonia	83,313 (9.2%)	86,490 (9.5%)
Urinary Tract Infection	109,092 (12.1%)	104,107 (11.4%)
Alzheimer’s Dementia	34,359 (3.8%)	27,991 (3.1%)
Malnutrition	38,755 (4.3%)	37,914 (4.2%)
Anxiety Disorder	157,260 (17.4%)	156,750 (17.2%)
Psychiatric Disorder other than Schizophrenia	13,670 (1.5%)	11,217 (1.2%)
Schizophrenia	11,487 (1.3%)	8.927 (1.0%)
Clinical Treatments
Intravenous Medication	58,138 (6.4%)	62,460 (6.8%)
Hemodialysis	23,877 (2.6%)	25,777 (2.8%)

^*Proportion of minutes in multiparticipant therapy sessions is calculated as the total group and concurrent therapy minutes divided by the total therapy minutes during each Skilled Nursing Facility stay.

Abbreviations: SD-Standard Deviation, OT- Occupational Therapy, PT- Physical Therapy, ADL- Activities of Daily Living, BIMS- Brief Interview for Mental Status

Analysis

We used generalized mixed effects logistic regression models with random effects for facility to account for clustering at the SNF level. We estimated the odds of improving in function and the odds of community discharge for patients receiving different levels of multiparticipant therapy, adjusted for patient and SNF stay covariates. As different patient risk factors are clinically relevant to each outcome, we adjusted for different patient covariates in the functional improvement and discharge models, guided by CMS risk adjustment methods used in the Nursing Home Compare quality measures program which found specific MDS variables to be associated with each outcome.²⁹ To avoid model overfitting and to adjust only for true confounders in the relationship between multiparticipant therapy and our outcomes, we further limited our adjustment to covariates that were found to be associated with multiparticipant therapy in a previous analysis.¹³ These adjustments also help reduce selection bias by accounting for potential clinical differences in patients who received different levels of multiparticipant therapy. Final covariates for each model are included in Tables 2 and 3.

Table 2.

Results of Mixed Effects Logistic Regression Model with Adjusted Odds Ratios for Patient Functional Improvement (N=901,544)

Proportion of Multiparticipant Therapy	Odds Ratio	95% CI	p-value
None (0%)	Ref.	Ref.	-
Low (>0 to <5%)	1.14	1.09–1.19	<0.001
Medium (5 to <25%)	1.18	1.13–1.24	<0.001
High (25% or more)	1.10	0.93–1.30	0.267
Covariate Effect Estimates
Demographics
Age (years)	0.99	0.99–0.99	<0.001
Female Sex	1.05	1.04–1.06	<0.001
Disability or ESRD as reason for Medicare Entitlement	0.75	0.53–1.05	0.093
Dual Eligibility	0.91	0.90–0.93	<0.001
SNF Stay Factors
Length of Stay (days)	1.03	1.03–1.03	<0.001
OT Minutes per Day of Therapy (15 min increase)	1.13	1.12–1.14	<0.001
PT Minutes Per Day of Therapy (15 min increase)	1.27	1.26–1.29	<0.001
Cognitive and Physical Function
ADL Scale Score on Admit (0–28)	1.01	1.01–1.01	<0.001
Cognitive Impairment
None (BIMS 13+)	Ref	Ref	-
Moderate (BIMS 8–12)	0.76	0.75–0.77	<0.001
Severe (BIMS 0–7)	0.51	0.50–0.52	<0.001
Active Diagnoses
Heart Failure	0.89	0.88–0.90	<0.001
Stroke	0.83	0.81–0.85	<0.001
Hip Fracture	1.12	1.10–1.14	<0.001
Other Fracture	1.07	1.06–1.09	<0.001

All covariates in the model are included above.

Abbreviations: CI-Confidence Interval, Ref-Reference, LOS- Length of Stay, OT- Occupational Therapy, PT- Physical Therapy, ADL- Activities of Daily Living, BIMS- Brief Interview for Mental Status

Table 3.

Results of Mixed Effects Logistic Regression Model with Adjusted Odds Ratios for Successful Community Discharge (N=912,996)

Proportion of Multiparticipant Therapy	Odds Ratio	95% CI	p-value
None (0%)	Ref.	Ref.	-
Low (>0 to <5%)	1.10	1.05–1.15	<0.001
Medium (5 to <25%)	1.44	1.34–1.55	<0.001
High (25% or more)	1.03	0.75–1.43	0.833
Covariate Effect Estimates
Demographics
Age (years)	1.00	1.00–1.00	<0.001
Female Sex	1.24	1.22–1.25	<0.001
Dual Eligibility	0.76	0.75–0.77	<0.001
SNF Stay Factors
Length of Stay (days)	1.04	1.04–1.04	<0.001
OT Minutes per Day of Therapy (15 min increase)	0.77	0.76–0.78	<0.001
PT Minutes Per Day of Therapy (15 min increase)	0.91	0.90–0.92	<0.001
Cognitive and Physical Function
ADL Scale Score on Admit (0–28)	0.84	0.84–0.84	<0.001
Cognitive Impairment
None (BIMS 13+)	Ref.	Ref.	-
Moderate (BIMS 8–12)	0.74	0.73–0.75	<0.001
Severe (BIMS 0–7)	0.71	0.70–0.72	<0.001
Needs Interpreter	1.30	1.25–1.35	<0.0001
Vision Impairment	0.91	0.89–0.92	<0.0001
Communication Impairment	0.81	0.79–0.84	<0.001
Falls in the last 6 months	1.00	0.99–1.01	0.782
Daily pain	0.81	0.77–0.85	<0.001
Use of an assistive device	1.33	1.29–1.38	<0.001
Wandering	1.00	0.95–1.04	0.842
Rejected Care	0.87	0.84–0.89	<0.001
Psychosis or Behavioral Symptoms	0.71	0.69–0.74	<0.001
Active Diagnoses
Stroke	0.85	0.84–0.87	<0.001
Hip Fracture	1.05	1.02–1.07	0.355
Wound Infection	0.82	0.77–0.87	<0.001
Paraplegia	0.72	0.64–0.80	<0.001
Multiple Sclerosis	1.15	1.05–1.26	0.002
Foot Infection	0.76	0.72–0.81	<0.001
Anemia	0.81	0.80–0.82	<0.001
Pneumonia	0.92	0.90–0.94	<0.001
Urinary Tract Infection	1.02	0.99–1.04	0.060
Alzheimer’s Dementia	1.20	1.16–1.24	<0.001
Malnutrition	0.69	0.67–0.71	<0.001
Anxiety Disorder	0.96	0.95–0.98	<0.001
Psychiatric Disorder other than Schizophrenia	0.90	0.86–0.95	<0.001
Schizophrenia	0.85	0.81–0.90	<0.001
Clinical Treatments
Intravenous Medication	0.66	0.65–0.68	<0.001
Dialysis	0.50	0.49–0.52	<0.001

All covariates in the model are included above. Abbreviations: CI-Confidence Interval, Ref-Reference, OT- Occupational Therapy, PT- Physical Therapy, ADL- Activities of Daily Living, BIMS- Brief Interview for Mental Status

Finally, we completed a sensitivity analysis exploring whether there were differential associations between outcomes and multiparticipant therapy provided by therapy disciplines separately. We ran one mixed effects logistic regression model for each outcome with separate independent variables for the overall proportions of multiparticipant PT and multiparticipant OT provided across SNF stays, adjusted for the same patient characteristics as our primary models and with a facility random effect. This study was approved by the University of Washington institutional review board. Analyses were conducted using statistical software (RStudio version 1.2.5019, R Foundation for Statistical Computing, Vienna Austria).

RESULTS

We identified over 1.1 million complete first SNF stays between January 1 and December 31, 2018 (Figure 1). Descriptive statistics for our samples, consisting of 901,544 patients with complete functional improvement data and 912,996 with complete discharge data, are included in Table 1. The vast majority of short-stay patients (97%) in SNFs did not receive multiparticipant therapy in SNFs in 2018. For the subset of patients receiving some multiparticipant therapy in each group, the median proportion of multiparticipant PT and OT was quite low at about 5%, and only 0.1% of patients received 25% or more multiparticipant therapy.

Figure 1:

The selection of Medicare fee-for-service Skilled Nursing Facility (SNF) stays from the Minimum Data Set 3.0 (MDS) during calendar year 2018. MBSF: Master Beneficiary Summary File, LOS: Length of Stay.

Descriptive statistics for outcomes and additional key covariates for each multiparticipant therapy group are provided in Supplemental Table 1. The results of the logistic regression analysis for the odds of improving in function among patients receiving different levels of multiparticipant therapy are presented in Table 2. Compared to no multiparticipant therapy, low and medium levels of multiparticipant therapy had small but statistically significant associations with positive outcomes; patients receiving up to 5% multiparticipant therapy had 14% higher odds of improving in function (95% confidence interval [CI]: 1.09–1.19) and patients receiving 5–25% multiparticipant therapy had 18% higher odds of functional improvement (95% CI 1.13–1.24). However, high multiparticipant therapy (25% or more) was not associated with differences in functional improvement compared to patients receiving all individualized therapy (95% CI 0.93–1.30).

The results of the model examining odds of discharging to the community for patients receiving different levels of multiparticipant therapy are presented in Table 3. Similarly, there was no statistically significant difference in community discharge rates for patients receiving high levels of multiparticipant therapy compared to no multiparticipant therapy. Compared to patients receiving only individualized therapy, low levels of multiparticipant therapy were associated with 10% higher odds of community discharge (95% CI 1.05–1.15) and medium levels were associated with 44% higher odds of community discharge (95% CI 1.34–1.55).

Our sensitivity analyses examining PT and OT disciplines separately found no association between the overall proportion of multiparticipant OT and functional improvement, but higher proportions of multiparticipant PT were associated with slightly higher odds of functional improvement (Supplementary Table 2). Effect sizes were small, with a 5% increase in multiparticipant PT associated with only 4% higher odds of functional improvement (95% CI: 1.02–1.05). When examining community discharge, multiparticipant OT and PT were both positively associated with improved community discharge rates. Five percent higher multiparticipant OT was associated with 2% higher odds of community discharge (95% CI 1.01–1.04), and five percent higher multiparticipant PT was associated with 5% higher odds of community discharge (95% CI 1.03–1.07).

DISCUSSION

This is the first analysis of a national cohort of SNF patients that examines associations between multiparticipant therapy and patient outcomes. Our adjusted results suggest that providing up to 25% of therapy minutes as multiparticipant therapy may help achieve small improvements in patient function and community discharge outcomes while potentially reducing overall therapy staffing costs.³⁴ However, positive associations between multiparticipant therapy minutes and both outcomes disappeared once the proportion of multiparticipant therapy reached 25% of total therapy time. Small sample sizes in the high multiparticipant therapy groups compared to the full cohort may have impacted the precision of this estimate, thus future work when post-PDPM data with higher multiparticipant therapy provision are available will be essential to confirm this finding. For now, the lack of benefit associated with high rates of multiparticipant therapy in our study supports 25% multiparticipant therapy time as a reasonable limit under PDPM, which was established in an effort to prioritize patient-centered care specific to individual patient needs.⁷ While high rates of multiparticipant therapy may be less costly for SNFs, until high rates of multiparticipant therapy are shown to be beneficial for patients, individualized therapy should remain the dominant modality of PT and OT treatment for the heterogenous SNF population that benefits from patient-centered care targeted at complex individual needs.³⁵

Specific to function, receiving both low and medium levels of multiparticipant therapy was associated with slightly higher functional improvement rates compared to no multiparticipant therapy in our adjusted model. These results are consistent with studies in international healthcare settings of multiparticipant therapy for specific patient populations, including a study by Vestri et al. that showed larger improvements in functional independence outcomes for patients with brain injury who received some multiparticipant therapy compared to those who received individual treatment only.²⁰ However, their study did not adjust for length of rehabilitation stay or overall intensity of therapy provision, and group therapy contents included structured activities that would fall outside of a typical physical or occupational therapy plan of care in the SNF setting, such as reading and socialization activities.²⁰ Another study by Renner et al. examined the efficacy of group versus individual progressive task training for patients after stroke during a six-week inpatient rehabilitation stay.¹⁹ The study randomly assigned patients to group versus individual treatment with dose-matched therapy intensity.¹⁹ The authors concluded that group therapy was as effective as individualized therapy for mobility outcomes, however, the long LOS and single diagnosis in this study may not be as generalizable to the SNF setting in the United States.¹⁹ Additionally, the relatively low level of baseline cognitive and functional impairments for patients in the Renner study highlights the ongoing need for individualized therapy, especially for patients with higher levels of functional or cognitive impairment who may not be able to as effectively participate in multiparticipant therapy sessions.^11,13,19

The potential positive impact of social support provided in a multiparticipant therapy setting may be a mechanism for the improvements in function seen for patients participating in low to medium amounts of multiparticipant therapy in this study. Benefits unique to the social environment in multiparticipant therapy sessions include decreased sense of social isolation and depression that could enhance motivation and improve participation and opportunities for peer feedback and vicarious learning.^18,36–38 However, lack of previous research on multiparticipant therapy in the heterogenous and complex SNF population, including lack of research on contents of multiparticipant therapy sessions, does not allow us to draw firm conclusions about the specific social benefits of multiparticipant therapy for the SNF population.

Similar to our functional improvement results, patients who received low to medium levels of multiparticipant therapy (up to 25% of total therapy time) had slightly higher odds of a positive community discharge outcome. This may be a novel result, as we are not aware of any other studies comparing individualized therapy to multiparticipant therapy for community discharge outcomes from any setting. The mechanism for potential improvements in community discharge rates for patients receiving low to medium levels of multiparticipant therapy in our cohort remains unknown, however, similar to functional improvement, social support, feedback, and vicarious learning during multiparticipant therapy sessions may allow patients to better achieve levels of independence for a safe discharge home. One qualitative study by Nilsson et al. evaluated a group occupational therapy program aimed at decreasing post-discharge challenges for older adults after a stroke, which emphasized reflection on the post-discharge environment, patients’ anticipated needs and abilities, and building capacity in household activities.³⁹ Participants in the group program reported increases in self-reflection, curiosity, and motivation that led to learning, anticipation, and goal-setting for home discharge. As there are no current studies examining specific contents of multiparticipant therapy session in SNFs, it is unknown whether multiparticipant therapy in the SNF setting similarly prepares patients for discharge.³⁹ Again, similar to functional improvement, it appears that spending ≥25% of therapy time in multiparticipant sessions may not be beneficial for community discharge. As specific discharge environments are unique across this heterogenous population, this result emphasizes the importance of individualized treatment to work towards safe community discharge for patients in SNFs.

Finally, the results of our sensitivity analysis that divided PT and OT disciplines suggest that only multiparticipant PT contributed to differences in functional improvement outcomes. This is unsurprising given the CMS definition of functional improvement used in this study, which includes only transfers, ambulation, and locomotion. OT treatment in SNFs emphasizes independence in self-care tasks and may not directly impact the mobility tasks measured in this study. Research using new outcome measures from assessment items introduced in SNFs in late 2018, which incorporate both self-care and mobility tasks, may allow for a more sensitive understanding of the impact of OT treatment on self-care function in SNFs.^40,41 For the sensitivity analysis examining community discharge, multiparticipant PT was a larger contributor to differences in community discharge rates than multiparticipant OT. Similar to the potential explanation that more therapy is provided for patients with worse discharge potential as discussed above, it may be that the self-care skills needed for successful community discharge addressed by OT require more individualization according to the patient discharge environment, while improving physical function necessary for safety and mobility in the home setting is more easily addressed through group exercises delivered by PTs.

Limitations

This cross-sectional study demonstrates associations between outcomes and multiparticipant therapy provision and does not imply direct causality. Additionally, while there is no established minimum clinically important difference for either outcome in this study, the effect sizes in this study for the association between multiparticipant therapy and functional improvement and discharge outcomes were small. However, similar or smaller effect sizes have been seen in other studies examining SNF rehabilitation outcomes,^11,27,42,43 and given this study’s large national sample of over 900,000 complete SNF stays in 2018, even a small increase in odds of a positive outcome may indicate a positive effect for a large number of patients.

Results are also generalizable only to Medicare short-stay patients admitted to SNF for rehabilitation. Low overall rates of multiparticipant therapy did not allow for a comparison of concurrent versus group therapy and outcomes, however, increases in the practice of multiparticipant therapy under PDPM may allow for the differentiation between the two modes of multiparticipant therapy provision in the future. Model limitations include the inability to adjust for upstream factors from the acute hospital stay that may be related to functional improvement or community discharge. Additionally, we carefully fit our models and adjusted for clustering at the facility level, but it is still possible that there is unmeasured confounding, such as differences in patient expectations for rehabilitation or other facility practices, that are unavailable on the MDS.¹¹ Further research should capitalize on increasing rates of multiparticipant therapy after PDPM to confirm these results and examine the mechanisms behind the impact of multiparticipant therapy on patient outcomes, to include examination of the content, social opportunities, and physiologic intensity of multiparticipant therapy sessions as well as the differential impact of multiparticipant therapy on different patient subgroups.

CONCLUSIONS

In the context of large increases in multiparticipant therapy provision in SNFs under PDPM, our results indicate that providing up to 25% of PT and OT in multiparticipant sessions is an efficient strategy of therapy delivery that may also provide benefit to patients in terms of functional improvement and discharge outcomes. However, multiparticipant therapy at or above the 25% threshold set by CMS was not positive for patient outcomes in our adjusted models. This research suggests that the majority of PT and OT treatment in SNFs should continue to be delivered through individualized therapy sessions despite removal of financial incentives for intensive individualized therapy under PDPM.

Clinical Implications.

1. Multiparticipant therapy in Skilled Nursing Facilities increased significantly under the Patient Driven Payment Model, but implications for patients are unknown.

2. Prior to the Patient Driven Payment Model, low and medium levels of multiparticipant physical and occupational therapy (up to 25% of total therapy minutes) were positively associated with small improvements in functional improvement and community discharge, but high levels (≥25% of total minutes) were not beneficial.

3. Providing up to 25% of physical and occupational therapy in multiparticipant sessions may be both efficient for staffing costs and provide small benefit for patients.