Is More Always Better? Financially motivated therapy and patient outcomes in Skilled Nursing Facilities

Abstract

Objective:

To determine if financially motivated therapy in Skilled Nursing Facilities (SNFs) is associated with patient outcomes.

Design:

Cohort study using 2018 Medicare administrative data.

Setting and Participants:

13,949 SNFs in the United States.

Participants:

934,677 Medicare Part A patients admitted to SNF for post-acute rehabilitation.

Interventions:

The primary independent variable was an indicator of financially motivated therapy, separate from intensive therapy, known as thresholding, defined as when SNFs provide ten or fewer minutes of therapy above weekly reimbursement thresholds.

Main Outcome Measures:

Dichotomous indicators of successful discharge to the community versus institution and functional improvement on measures of transfers, ambulation, or locomotion. Mixed effects models estimated relationships between thresholding and community discharge and functional improvement, adjusted for therapy intensity, patient, and facility characteristics. Sensitivity analyses estimated associations between thresholding and outcomes when patients were stratified by therapy volume.

Results:

Thresholding was associated with a small positive effect on functional improvement (odds ratio [OR] 1.07; 95% CI 1.06–1.09) and community discharge (OR 1.03, 95% CI 1.02–1.05). Effect sizes for functional improvement were consistent across patients receiving different volumes of therapy. However, effect sizes for community discharge were largest for patients in low-volume therapy groups (OR 1.27, 95% CI 1.18–1.35).

Conclusions:

Patients who experienced thresholding during post-acute SNF stays were slightly more likely to improve in function and successfully discharge to the community, especially for patients receiving lower volumes of therapy. While thresholding is an inefficient and financially motivated practice, results suggest that even small amounts of extra therapy time may have contributed positively to outcomes for patients receiving lower-volume therapy. As therapy volumes decline in SNFs, these results emphasize the importance of Medicare payment policy designed to promote, not disincentivize, potentially beneficial rehabilitation services for patients.

Twenty percent of all hospitalized Medicare beneficiaries receive post-acute care in skilled nursing facilities (SNFs),¹ with SNF costs for the Centers for Medicare and Medicaid Services (CMS) totaling over $28 billion in 2020.² In October 2019, the Centers for Medicare and Medicaid Services (CMS) drastically altered payment policy for fee-for-service Part A patients in response to rising costs and resource utilization in SNFs.¹ Specifically, rising physical therapy (PT) and occupational therapy (OT) intensity in SNFs was considered by CMS to be financially motivated rather than clinically indicated.^1,3–5 Under the previous Resource Utilization Group (RUG-IV) payment system, SNF reimbursement for Medicare fee-for-service patients was based on hours of weekly therapy provided, and reimbursement rates increased at specific intervals of therapy time per week.⁶ Therapy RUG-IV groups ranged from Low (45–149 minutes of combined PT, OT, and speech therapy per week) to Ultrahigh (720+ minutes per week), with Ultrahigh therapy garnering the highest reimbursement.⁶ In addition to incentivizing higher therapy volumes overall, this system also incentivized thresholding behavior, defined by CMS as a practice of providing ten or fewer minutes of therapy above the weekly threshold for higher reimbursement in order to garner higher payment with minimal increase in therapy staffing costs.^6–11 Sharp increases in both overall therapy intensity and thresholding behavior in SNFs in the early 2000s motivated CMS to develop the Patient Driven Payment Model (PDPM), which, in part, aims to decrease potentially unnecessary PT and OT in SNFs by removing higher payments for higher volumes of therapy and aligning payment with patient characteristics.¹²

While PDPM intended to reduce financially motivated therapy, the design of PDPM disincentivizes intensive therapy across the board without incorporating rehabilitation outcomes in payment determinations.^10,13 Multiple studies have linked intensive therapy with improved patient outcomes in SNFs,¹⁴ thus, if high-volume therapy may be clinically indicated, using overall therapy minutes to reflect financially-motivated practice would not be accurate. However, the optimal therapy intensity to achieve desired patient outcomes in SNFs is unknown, especially because the RUG-IV system led to minimal variation in therapy intensity as SNFs sought to place patients into more profitable payment groups.^1,15 Without the ability to distinguish between clinically indicated and financially motivated therapy provision, rehabilitation payment policy will likely continue to be driven by costs rather than quality.¹⁶

Monitoring thresholding behavior addresses this gap by providing an indicator that allows for specific study of financially motivated therapy. Thresholding occurs when patients receive a certain amount of therapy close to each specific threshold for higher reimbursement, even for patients receiving low therapy volumes.¹⁰ Thresholding was common prior to PDPM, and an average of 62% of patients receiving Ultrahigh therapy experienced thresholding.¹⁰ Facilities that engaged in thresholding to move a patient into a higher paying RUG-IV group could experience a return of hundreds of dollars per patient per week for a minimal investment in terms of services provided.⁹ Thus, thresholding is a unique indicator of financially motivated therapy that can be examined separately from intensive therapy, which could be clinically indicated.

Previous work analyzing a national cohort of SNFs showed that SNF characteristics such as staffing processes and for-profit ownership were associated with higher rates of thresholding behavior for patients receiving Ultrahigh intensity therapy.¹⁰ Thresholding was also associated with worse outcomes at the facility level, including lower rates of community discharges and higher rates of 30-day hospital readmissions, but effect sizes were very small.¹⁰ Another facility-level analysis found that SNFs with high rates of thresholding into the Ultrahigh therapy intensity payment category had higher 30-day hospital readmission rates, longer lengths of stay, and higher costs per SNF stay.¹¹ However, there currently is no evidence on relationships between thresholding and outcomes at the patient level. Additionally, associations between thresholding and outcomes for patients receiving volumes of therapy besides Ultrahigh intensity therapy are unknown. The purpose of this study is to use patient-level data to isolate the relationship between financially motivated therapy – operationalized as thresholding – and patient outcomes, while controlling for overall therapy intensity.

Methods

Data Sources and Study Population

We created a national cohort of Medicare Part A fee-for-service SNF stays using the 2018 Minimum Dataset (MDS) 3.0 and the 2018 Medicare Beneficiary Summary File (MBSF). The MDS is a required comprehensive assessment for all patients in SNFs and includes demographic, functional, and clinical patient information as well as data on treatments received during SNF admission, including therapy services.^17,18 The MBSF includes demographic and Medicare coverage information. We created complete SNF stays by including admission, discharge, and all interim MDS assessments, and then selected the first complete stay per patient in 2018 to avoid inducing correlation by using multiple stays per patient.¹⁹ To create our cohort of short-stay Medicare patients admitted to SNF for post-acute rehabilitation, we excluded patients missing insurance eligibility information from the MBSF file, long-stay patients with SNF length of stay (LOS) over 100 days, patients admitted and discharged on the same day, patients admitted to SNF for hospice care, patients who died during SNF admission, patients who were comatose on admission, and patients who did not receive physical or occupational therapy during their stay.^13,15,20,21 See supplement for cohort creation flowsheet.

Independent Variable

The primary independent variable of interest was a dichotomous indicator for whether patients experienced thresholding in any assessment period during their SNF stay. Thresholding occurred if the total therapy minutes for the assessment period exceeded the minute threshold for the assigned RUG-IV category by ten or fewer minutes.^4,12 This definition of thresholding was set by CMS based on distributions of therapy minutes around payment thresholds.⁴ Thresholding was calculated from the MDS, which reports the number of physical, occupational, and speech therapy minutes in the last seven days of each assessment period as well as the RUG-IV category which determined payment for that assessment. Combined therapy minutes for each assessment period were summed. Using the methods used by CMS to calculate total therapy minutes under the RUG-IV system, we divided any therapy minutes categorized as group therapy by four and minutes categorized as concurrent therapy by two before adding group or concurrent therapy minutes to the total.²²

Outcome Variables

The outcomes of interest were successful discharge to the community and patient functional improvement. Functional improvement occurred when scores on MDS transfer, ambulation, or locomotion items improved between admission and discharge.^23,24 This dichotomous measure of functional improvement was used by CMS to monitor SNF quality of care for short-stay patients prior to PDPM.²⁴ Community discharge was considered successful if the MDS indicated discharge to a noninstitutional setting.

Covariates

To isolate the effects of financially motivated thresholding from potentially clinically indicated therapy volume, we adjusted for the overall intensity of therapy during the full SNF stay, operationalized as average minutes of occupational and physical therapy per day of therapy received during the SNF stay. Other covariates from the MBSF or admission MDS assessment included demographics: age, sex, disability or end-stage renal disease as current reason for Medicare entitlement, an indicator for dual Medicare-Medicaid eligibility operationalized as either full or partial dual status in the month of SNF admission, a dichotomous indicator of marital status, and need for an interpreter. We included indicators from the MDS admission assessment for vision and communication impairments, and cognitive impairment scores from the Brief Interview for Mental Status, categorized as no cognitive impairment (scores of 13 or higher), moderate cognitive impairment (scores of 8–12), or severe cognitive impairment (scores under 8).^20,25 We adjusted for activity of daily living scale scores at admission, which ranges from 0–28, with higher scores indicating more severe impairment on seven functional tasks.^26,27 We included indicators for falls within the last six months, use of an assistive device, delirium, major depression, daily reported pain, wandering, rejection of care, and psychosis such as hallucinations or physical or verbal behavioral symptoms directed towards others.^18,23 To reflect medical complexity we included indicators for active medical diagnoses on SNF admission and clinical treatments received during the SNF stay, included in Table 1.

Table 1.

Descriptive statistics for complete short-stay skilled nursing facility patient stays in 2018 (n=934,677)

	Patients experiencing thresholding	Patients not experiencing thresholding
	Mean (SD) or N (%)	Mean (SD) or N (%)	p-value
Number of Patients	527,196 (56.4%)	407,481 (43.6%)
Outcomes
Improved in Function	295,371 (64.08%)	176,061 (46.79%)	<0.001
Successful Community Discharge	355,819 (77.99%)	268,046 (71.75%)	<0.001
Demographics
Age (years)	80.71 (8.51)	80.25 (8.50)	<0.001
Female	325,867 (61.81%)	244,527 (60.01%)	<0.001
Disability or ESRD as reason for Medicare Entitlement	113 (0.02%)	87 (0.02%)	0.98
Dual Eligibility	106,478 (20.20%)	64,241 (15.77%)	<0.001
Married	178,748 (33.91%)	153,954 (37.78%)	<0.001
Needs Interpreter	15,172 (2.88%)	9,528 (2.34%)	<0.001
SNF Stay Characteristics
Length of Stay (days)	30.02 (21.00)	18.88 (16.49)	<0.001
PT Minutes per Day	57.24 (10.70)	54.89 (12.35)	<0.001
OT Minutes per Day	56.05 (10.36)	53.49 (11.70)	<0.001
Cognitive and Physical Function
ADL Scale Score on Admit (0–28)*	17.07 (4.02)	16.59 (4.51)	<0.001
Falls in the last 6 months	249,695 (47.36%)	176,967 (43.43%)	<0.001
Daily pain	6,773 (1.28%)	6,202 (1.52%)	<0.001
Use of an assistive device	507,563 (96.29%)	388,194 (95.31%)	<0.001
Cognitive Impairment
None (BIMS 13+)	330,908 (65.61%)	248,787 (68.92%)	<0.001
Moderate (BIMS 8–12)	104,566 (20.73%)	69,148 (19.16%)	<0.001
Severe (BIMS 0–7)	68,896 (13.66%)	43,035 (11.92%)	<0.001
Vision Impairment	80.797 (15.35%)	61,008 (14.97%)	<0.001
Communication Impairment	29,057 (5.51%)	22,653 (5.56%)	0.32
Wandering	9,336 (1.77%)	7,359 (1.81%)	0.20
Rejected Care	22,282 (4.23%)	19,983 (4.90%)	<0.001
Active Diagnoses **
Stroke	41,118 (7.99%)	32,060 (7.87%)	0.03
Hip Fracture	49,916 (9.47%)	32,263 (7.92%)	<0.001
Other Fracture	66,602 (12.63%)	44,245 (10.86%)	<0.001
Cancer	44,976 (8.53%)	26,650 (6.54%)	<0.001
Diabetes Mellitus	173,933 (32.99%)	129,084 (31.68%)	<0.001
Wound Infection	4,075 (0.77%)	3,144 (0.77%)	0.94
Foot Infection	5,271 (1.00%)	3,638 (0.89%)	<0.001
Anemia	143,833 (27.28%)	107,313 (26.34%)	<0.001
Asthma, COPD, or Chronic Lung Disease	128,702 (24.41%)	99,935 (24.53%)	0.21
Pneumonia	48,505 (9.20%)	39,246 (9.63%)	<0.001
Heart Failure	123,720 (23.47%)	93,123 (22.86%)	<0.001
Hypertension	416,844 (79.07%)	313,434 (76.92%)	<0.001
Septicemia	22,002 (4.17%)	16,888 (4.14%)	0.49
Urinary Tract Infection	64,686 (12.27%)	47,015 (11.54%)	<0.001
Alzheimer’s Dementia	21,432 (4.07%)	13,909 (3.41%)	<0.001
Dementia Other than Alzheimer’s	95,439 (18.10%)	62,920 (15.44%)	<0.001
Parkinson’s Disease	20,904 (3.97%)	15,415 (3.78%)	<0.001
Malnutrition	23,516 (4.46%)	17,230 (4.23%)	<0.001
Psychosis or Behavioral Symptoms	23,933 (4.54%)	19.672 (4.83%)	<0.001
Major Depression	25,093 (4.76%)	18,765 (4.61%)	<0.001
Anxiety Disorder	92,753 (17.59%)	70,606 (17.33%)	<0.001
Psychiatric Disorder other than Schizophrenia	8,920 (1.69%)	5,422 (1.33%)	<0.001
Schizophrenia	7,555 (1.45%)	4,227 (1.04%)	<0.001
Clinical Treatments
Intravenous Medication	33,289 (6.31%)	27,350 (6.71%)	<0.001
Hemodialysis	15,096 (2.86%)	10,399 (2.55%)	<0.001
Facility Characteristics
Urban	436,159 (82.73%)	347,179 (85.20%)	<0.001
For-Profit	372,790 (70.72%)	257,253 (63.13%)	<0.001
Hospital-Based	12,961 (2.46%)	30,193 (7.41%)	<0.001
Chain Affiliation	316,043 (59.95%)	240,922 (59.12%)	<0.001
Contractor Staffing
All In-House Staff	185,301 (35.15%)	181,122 (44.45%)	<0.001
All Contract Staff	281,566 (53.41%)	177,403 (43.54%)	<0.001
Mix	60,329 (11.44%)	48,956 (12.01%)	<0.001
Assistant Staffing
None	34,702 (6.58%)	36,305 (8.91%)	<0.001
0– <25%	86,002 (16.31%)	76,604 (18.80%)	<0.001
25– <50%	129,473 (24.46%)	104,163 (25.56%)	<0.001
50– <75%	135,481 (25.70%)	99,447 (24.41%)	<0.001
75%+	141,538 (26.85%)	90,962 (22.32%)	<0.001

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

^*Higher scores on the ADL Scale score indicate more severe functional impairment.

^**These are not mutually exclusive diagnosis groups because patients may have more than one active diagnosis coded on the admission assessment.

Finally, we controlled for specific organizational characteristics that have been shown to be associated with quality outcomes and/or thresholding behavior using calendar year 2018 LTCFocus data²⁸ and the CMS Provider of Services Files.^10,29–31 We included urban versus rural county location, a dichotomous indicator of for-profit ownership, hospital-based versus freestanding location, chain affiliation, contractor therapy staffing (characterized as 100% in-house PT and OT staff, 100% contractors, or a mix), and assistant staffing, characterized as 0% PT and OT assistants, and then as quartiles of assistant staffing for SNFs staffing any assistants.

Analyses

In our primary analysis, we estimated relationships between thresholding and odds of functional improvement and community discharge using generalized mixed effects logistic regression models with a random intercept for facility to control for correlation between patients admitted to the same SNF due to unobserved SNF-level factors. Both models adjusted for SNF LOS, therapy volume, patient demographics, ADL scores on admission, and cognitive scores. To avoid model overfitting, each model included only the active diagnoses on admission that CMS had previously found to be associated with either the functional improvement or community discharge outcome.²³ Specifically, the community discharge model adjusted for all active diagnoses in Table 1, while the functional improvement model adjusted only for diagnoses associated with that outcome: heart failure, stroke, hip fracture, and other fracture.

We then ran sensitivity analyses to explore whether thresholding was similarly associated with outcomes across SNF stays with different volumes of average weekly therapy. We hypothesized that patients receiving overall lower volumes of therapy may experience more benefit from the additional few minutes of therapy services per week received when thresholding occurs compared to patients already receiving high levels of therapy. For sensitivity analyses, we stratified patients into three groups based on their average combined PT and OT minutes per week throughout their SNF stay: Ultrahigh therapy group (patients receiving at least 720 minutes of weekly therapy), Very High therapy group (patients receiving between 500–720 minutes of weekly therapy), and Other therapy (combining all patients receiving under 500 minutes of therapy per week due to small sample size). We then ran logistic regression models with facility random effects for functional improvement and discharge outcomes for the three groups. Models for sensitivity analyses included all the same covariates as the primary analyses, except for therapy minutes per day, as patients were already stratified based on therapy volume. Analyses were conducted using statistical software (RStudio version 1.2.5019, R Foundation for Statistical Computing, Vienna Austria). This study was approved by the institutional review board of University of Washington.

Results

Descriptive statistics for outcomes, patient demographic and clinical characteristics, and SNF stay characteristics for the 934,677 SNF stays in 13,949 SNFs that met inclusion criteria are included in Table 1. Thresholding was common in our sample, with 56.4% of patients experiencing thresholding during their SNF stay.

Results of our primary analyses are included in Table 2. Thresholding was associated with a small positive effect on functional improvement, with patients who experienced thresholding having 1.07 times higher odds of improving in function (95% CI 1.06, 1.09). This odds ratio equated to a 1.5% higher predicted probability of improving in function for patients who experienced thresholding (average predicted probability 66.62% for patients who experienced thresholding versus 65.12% for those who did not experience thresholding). Similarly, thresholding was associated with 1.03 times higher odds of discharging to the community (95% CI 1.02, 1.05). This odds ratio equated to a 0.5% higher predicted probability of discharging to the community for patients who experienced thresholding (average predicted probability 78.54% for patients who experienced thresholding versus 78.04% for those who did not experience thresholding).

Table 2.

Adjusted* associations between functional improvement and community discharge outcomes and thresholding.

Variable	Odds Ratio (95% CI)	Marginal Probability**	p-value
Functional Improvement	1.07 (1.06, 1.09)	1.50%	<0.001
Community Discharge	1.03 (1.02, 1.05)	0.50%	<0.001

Results are from logistic mixed effects regression models that were adjusted for patient demographics, Skilled Nursing Facility length of stay, therapy volume, functional scores on admission, cognition scores, and facility characteristics. Different patient diagnoses were selected for functional improvement and community discharge models to avoid overfitting according to Medicare risk adjustment methods. Community discharge models included all active diagnoses noted in Table 1, and functional improvement models included only those previously associated with discharge function: heart failure, stroke, hip fracture, and other fracture.²³

^**Marginal probabilities reflect differences in predicted probabilities of each outcome for patients experiencing thresholding versus those who did not.

Results from sensitivity analyses stratifying patients based on volume of therapy are included in Table 3. Results for relationships between thresholding and functional improvement stratified by therapy volume were similar to the primary analysis and similar across therapy volume groups. In contrast, consistent with our hypothesis, patients receiving <500 therapy minutes per week had larger positive effect sizes for community discharge outcomes if they experienced thresholding. Among patients receiving ultrahigh PT and OT, patients who experienced thresholding had 1.07 times higher odds (0.86% higher probability) of community discharge compared to those who did not experience thresholding. Among patients receiving <500 minutes per week of therapy, patients who experienced thresholding had 1.27 times higher odds (4.16% higher probability) of community discharge compared to those who did not experience thresholding.

Table 3.

Adjusted^* associations between thresholding and outcomes for therapy volume subgroups.

	Functional Improvement
Therapy Volume Subgroup	Odds Ratio (95% CI)	Marginal Probability**	p-value
Ultrahigh (n =543,973)	1.10 (1.09, 1.12)	1.92%	<0.001
Very High (n = 247,071)	1.07 (1.05, 1.10)	1.56%	<0.001
Other (n = 32,216)	1.10 (1.04, 1.18)	2.27%	0.003
	Community Discharge
Therapy Volume Subgroup	Odds Ratio (95% CI)	Marginal Probability	p-value
Ultrahigh (n =543,973)	1.07 (1.05, 1.08)	0.86%	<0.001
Very High (n = 247,071)	1.03 (1.01, 1.06)	0.63%	0.002
Other (n = 32,216)	1.27 (1.18, 1.35)	4.16%	<0.001

^*Results are from logistic mixed effects regression models that were adjusted for patient demographics, Skilled Nursing Facility length of stay, therapy volume, functional scores on admission, cognition scores, and facility characteristics. Different patient diagnoses were selected for functional improvement and community discharge models to avoid overfitting according to Medicare risk adjustment methods. Community discharge models included all active diagnoses noted in Table 1, and functional improvement models included only those previously associated with discharge function: heart failure, stroke, hip fracture, and other fracture.²³

^**Marginal probabilities reflect differences in predicted probabilities of each outcome for patients experiencing thresholding versus those who did not.

Discussion

In this study, we found small positive associations between patient outcomes and receiving therapy just over weekly minute thresholds for higher reimbursement in SNFs, when adjusting for overall therapy intensity and patient and facility characteristics. Small positive associations between thresholding and functional improvement were consistent for patients receiving different volumes of therapy during their SNF stay. Effect sizes for the relationship between thresholding and successful community discharge were highest for patients receiving under 500 minutes of therapy per week.

Results of this patient-level analysis suggest that the financially motivated therapy billing practice of thresholding, while inefficient for CMS as the payer, is associated with positive outcomes for individual patients.¹⁰ The small positive effect sizes may reflect that the extra minutes of therapy provided for patients experiencing thresholding constitute a ‘bonus’ in their overall therapy time beyond what would typically have been provided. Especially for patients receiving low volumes of weekly therapy, extra time spent in therapy sessions may provide additional time for planning and education, safety training, and skills improvement needed for successful community discharge.

In addition to extra time in therapy for patients experiencing thresholding, higher therapy intensity overall also had positive effect sizes for both outcomes, with 10 minutes more OT or PT per day associated with higher odds of functional improvement and community discharge (Supplemental Tables 1 and 2). This result is consistent with a growing body of evidence on positive relationships between intensive therapy and outcomes for patients in SNFs.^13–15,32 Thus these results suggest that, even when financially motivated, higher volumes of therapy provision may help SNFs aiming to provide quality care and improve patient outcomes, especially important as CMS moves towards value-based payment initiatives with plans to specifically incorporate rehabilitation-related outcomes such as community discharge and function in the SNF value-based purchasing program.³³ These results are also concerning in the light of reports of declining therapy provision in SNFs under PDPM and during the COVID-19 pandemic.^34–37

Positive effects of thresholding in this analysis were inconsistent with previous research that found small negative associations between thresholding and patient outcomes at the facility level.¹⁰ The facility-level analysis used functional improvement, discharge, and readmission outcome measures that were risk-adjusted for patient clinical factors associated with each outcome for each SNF. However, the effect sizes in the facility-level analysis were so small as to indicate a clinically insignificant relationship between thresholding and patient outcomes. Differences between facility- and patient-level analyses point to the need for analyses of individual patient outcomes in addition to facility-level costs and quality under PDPM. Together, these two studies support a conclusion that thresholding may be an inefficient financially motivated billing practice, but additional therapy may contribute to positive outcomes for individual patients.

While thresholding behavior is no longer incentivized in SNFs under PDPM because of the removal of the RUG-IV system, SNFs have historically responded to changes in payment policy incentives by altering clinical and administrative processes.^38–40 Thus, even as CMS continues to move towards more patient-centered and value-based payment policy, the industry will likely continue to respond in ways to maximize profits, as seen in reports of decreased therapy staffing and therapy provision under PDPM.^34–37,41 In that context, these results reinforce the importance of careful payment policy design that does not disincentivize services that may be beneficial to patients.

Limitations

This observational study demonstrates associations between thresholding behavior and patient outcomes and does not imply causality. Hospital readmission outcomes have previously been examined in the context of thresholding behavior at the facility level,¹⁰ but individual readmission outcomes could not be calculated from MDS data. Future work incorporating claims data could be utilized to examine relationships between thresholding and readmissions. Additionally, MDS data do not allow for a complete picture of care trajectories, including upstream factors from hospitalization, inpatient rehabilitation admissions preceding SNF stays, or hospitalizations interrupting SNF stays, so we were unable to directly adjust for factors from hospital stays that may be related to function or discharge outcomes. Finally, we examined associations between thresholding and outcomes across the heterogenous SNF population in 2018, when all active diagnoses were documented on the MDS with no ability to differentiate acuity or severity, which may have masked the size or direction of effects for specific patient groups, such as acute fracture or stroke. Primary diagnosis for SNF stay was added to the MDS after PDPM, which will allow for the examination of differential policy impacts on specific diagnostic subgroups.

Speech therapy was included in total minute calculations to detect thresholding, as weekly physical, occupational, and speech minutes are totaled to determine RUG group and indicate how close a patient was to a payment threshold. However, speech therapy minutes were not included in therapy intensity adjustments or sensitivity analysis stratifications because speech therapy costs have been shown to be inversely related to physical and occupational therapy costs prior to PDPM.⁴ Because patients receiving high-intensity physical and occupational therapy did not typically receive high-intensity speech therapy prior to PDPM, we would not expect speech therapy minutes to be utilized to achieve thresholding in the same manner as physical and occupational therapy minutes. However, future analyses could include all therapy disciplines to understand outcomes of all therapy services.

Conclusions

This study found that patients undergoing post-acute rehabilitation in SNFs who experienced thresholding behavior were slightly more likely to improve in function and successfully discharge to the community, especially for patients receiving lower volumes of therapy. While thresholding is a financially motivated inefficient billing practice that is disincentivized under new payment policy, it appears that additional therapy minutes may have even contributed positively to patient outcomes. As therapy volumes decline under PDPM, these results emphasize the importance of Medicare payment policy designed to promote, not disincentivize, beneficial rehabilitation services for patients.

Abbreviations:

SNF

Skilled Nursing Facility

OR

odds ratio

CMS

Centers for Medicare & Medicaid Services

RUG-IV

Resource Utilization Group

PT

physical therapy

OT

occupational therapy

PDPM

Patient Driven Payment Model

MDS

Minimum Data Set

MBSF

Master Beneficiary Summary File

LOS

length of stay

ADL

Activities of Daily Living