Skip to main content
White paper

Variation in skilled nursing facility practice patterns: Opportunities exist for more efficient management

13 September 2024

There is significant variation in skilled nursing facility (SNF) average length of stay (ALOS) and readmission rates among Medicare fee-for-service (FFS) beneficiaries admitted to a SNF following an acute inpatient hospital stay. These variations are unexplained by differences in patient case mix and highlight the opportunity for more efficient management of SNF stays.

Skilled nursing facility services represent a significant proportion of Medicare FFS expenditures. A 2024 Medicare Payment Advisory Commission (MedPAC) report noted that, in 2022, about 14,700 SNFs furnished approximately 1.8 million Medicare-covered stays to 1.3 million FFS beneficiaries, which amounted to $29 billion in spending on SNF services.1 This represents approximately 6.3% of total Medicare FFS spending based on our analysis of the Medicare 2022 100% FFS data.

Reducing medically unnecessary transfers to SNFs after acute inpatient stays, as well as reducing medically unnecessary days during a SNF stay and medically unnecessary readmissions from SNF back to hospital, can have a substantial impact on Medicare costs. Medicare Advantage (MA) plans, Medicare accountable care organizations (ACOs), and Medicare bundled payment participants can evaluate the efficiency of their SNF providers by comparing performance on risk-adjusted ALOS and readmission rates among SNFs. Findings can provide direction for effective SNF contracting strategies.

Do patterns in length of stay across all SNF cases suggest opportunity?

To establish whether opportunity exists for more efficient length of stay (LOS) management among SNF stays, we used the Medicare 100% Research Identifiable Files (RIF) FFS data to evaluate patterns in SNF LOS. We evaluated SNF stays that were admitted between September 22, 2022, and September 22, 2023. To allow for capture of SNF stays lasting up to 100 days we looked at data through December 31, 2023. We used Centers for Medicare and Medicaid Services (CMS) SNF Certification Numbers (CCNs) to identify SNFs and we excluded transitional care facilities, acute inpatient rehab facilities, and other non-SNF rehabilitation facilities. To minimize inclusion of facilities that may not be considered a SNF, we excluded SNFs where 75% or more of their total stays had a LOS less than or equal to 14 days. Our final sample included 627,095 SNF stays among 5,545 total SNFs.

In Figure 1, we plot the LOS of each individual SNF stay and identify a pattern for SNF LOS that correlates with Medicare’s SNF benefit coverage policy. There is an observable spike in LOS at 20 days, which corresponds to beneficiary cost sharing starting at day 21. Medicare pays 100% of SNF stay costs for the first 20 days, and then beneficiaries pay $204 per day (the 2024 rate) for days 21 to 100.2 There is another spike in LOS at 100 days, when beneficiaries become 100% responsible for SNF payment. We also observe spikes at 7 days, 14 days, and 28 days, corresponding to common SNF practice patterns that evaluate the need for continued stay on a weekly basis. These findings could indicate an opportunity for more efficient management of SNF LOS based on more frequent evaluation of the need for continued stay.

Figure 1: Distribution of SNF admissions by length of stay (Days 1-100)

FIGURE 1: DISTRIBUTION OF SNF ADMISSIONS BY LENGTH OF STAY  (DAYS 1-100)

Source: Milliman analysis of 9/22/2022-9/22/2023 Medicare 100% RIF data. Analysis based on 627,095 SNF stays among 5,545 total SNFs.

Is there variation in length of stay and readmission rates by unique SNF?

In order to investigate variation in performance by SNF, we used our sample of 627,095 SNF stays among 5,545 total SNFs and assigned each SNF admission to a unique SNF. We calculated ALOS across all stays for each SNF. Because some SNFs may have lower ALOS simply because they send patients back to hospital at a higher-than-average rate, we also calculated the average readmission rate back to hospital within one day of SNF discharge for each SNF. When comparing SNF performance, both metrics are essential data elements.

Identifying efficient SNF management by simply comparing raw ALOS and readmission rates among SNFs may not account for differences in the case mix of patients admitted to each SNF. In order to compare ALOS and readmission rates among SNFs, risk adjustment is necessary to control for relevant differences in the characteristics of a SNF’s patient population. We developed separate risk adjustment models for LOS and readmissions to more credibly compare ALOS and readmission rates between SNFs. We constructed a multiple linear regression model to estimate the expected SNF LOS for a given SNF stay and a multiple logistic regression model to estimate the expected chance of a readmission following a given SNF stay based on the complexity of patients at the time of their SNF admission. The patient characteristics that were considered potentially influential on SNF ALOS and readmission rates were used as model inputs and are shown in the table in Figure 2.

Figure 2: Model input summary

MODEL INPUT CATEGORY MODEL INPUT
Patient Demographics Sex
Age at SNF Admission
Dual Eligibility Status
Race
Institutional Status
Patient Health Status Alzheimer’s Status
Cognitive Disabilities Indication
CMS HCC Risk Score (V. 28)
Charlson Comorbidity Categories
Count of DME Claims in 360 Days Prior
Count of Inpatient Admits in 90 Days Prior
Count of Hospice Claims in 90 Days Prior
Preceding Inpatient Stay Surgical Admission Indicator
Hospital Teaching Status
DRG
Length of Stay
Admission Type
Number of Days in ICU

To balance model accuracy and simplicity, the final models were chosen using a backward stepwise regression procedure that selects the final model based on the lowest value of the Akaike information criterion (AIC). This procedure takes all model inputs and builds subsequent regression models by iteratively fitting models that remove the weakest model input, ultimately choosing a final set of model inputs that balances the risk of overfitting against the risk of underfitting.

Findings

In Figures 3 and 4 we examine variation in SNF risk-adjusted ALOS and readmission rates, respectively. In both figures we see a significant distribution of facilities below and above the most frequent finding, indicating wide variation in these metrics from facility to facility, even after accounting for case-mix differences with risk adjustment.

Figure 3: Distribution of risk adjusted average length of stay by SNF

FIGURE 3: DISTRIBUTION OF RISK ADJUSTED AVERAGE LENGTH OF  STAY BY SNF

Source: Milliman analysis of 9/22/2022-9/22/2023 Medicare 100% RIF data. Analysis based on 627,095 total SNF admissions among 5,545 total SNFs.

Figure 4: Distribution of risk adjusted readmission rate by SNF

FIGURE 4: DISTRIBUTION OF RISK ADJUSTED READMISSION RATE BY SNF

Source: Milliman analysis of 9/22/2022-9/22/2023 Medicare 100% RIF data. Analysis based on 627,095 total SNF admissions among 5,545 total SNFs.

To examine geographic variation in SNF performance, we assigned each SNF to its metropolitan statistical area (MSA) based on the SNF’s physical location reported by CMS. The table in Figure 5 summarizes risk-adjusted (to represent a nationwide average mix of SNF patient severity) ALOS and readmission rates for the 15 highest-volume MSAs. Among these 15 MSAs, SNF ALOS ranges from 20.4 to 35.3 days, and readmission rates range from 14.1% to 35.3%.

Figure 5: SNF performance for top 15 MSAs with highest annual SNF admits

MSA SNF
ADMITS
RISK-
ADJUSTED
SNF ALOS
RISK-
ADJUSTED
READMIT RATE
New York-Jersey City-White Plains, NY-NJ 23,709 33.4 30.0%
Chicago-Naperville-Evanston, IL 21,557 24.6 23.3%
Nassau County-Suffolk County, NY 15,138 34.7 24.3%
Los Angeles-Long Beach-Glendale, CA 11,409 32.0 35.8%
New Brunswick-Lakewood, NJ 11,386 26.0 23.1%
Baltimore-Columbia-Towson, MD 11,182 26.2 20.0%
Cambridge-Newton-Framingham, MA 9,839 20.7 18.8%
Phoenix-Mesa-Chandler, AZ 9,483 20.9 13.4%
Washington-Arlington-Alexandria, DC-VA-MD-WV 9,052 25.3 21.7%
Tampa-St. Petersburg-Clearwater, FL 8,891 22.9 23.0%
Boston, MA 8,719 21.0 17.7%
Warren-Troy-Farmington Hills, MI 8,373 21.6 19.5%
Montgomery County-Bucks County-Chester County, PA 8,296 21.0 19.1%
Newark, NJ-PA 7,245 25.4 21.8%
West Palm Beach-Boca Raton-Boynton Beach, FL 6,824 22.6 21.3%

Source: Milliman analysis of 9/22/2022-9/22/2023 Medicare 100% RIF data. Analysis based on 627,095 SNF stays among 5,545 total SNFs.

In the table in Figure 6, we summarize variation by facility within one MSA: Chicago-Naperville-Evanston, IL, a large MSA with roughly average LOS and readmission rates. We show risk-adjusted ALOS and readmission rates for the 25 highest-volume SNFs. These results show significant variation in performance among SNFs within the same MSA. ALOS ranges from 17.8 to 31.1 days, and readmission rates range from 8.6% to 29.7%.

Figure 6: Performance for top 20 SNFs by admits in Chicago-Naperville-Evanston, IL MSA

SNF # # SNF
ADMITS
RISK-ADJUSTED
SNF ALOS
RISK-ADJUSTED
READMISSION RATE
1 1,091 18.4 10.8%
2 564 25.9 11.7%
3 460 23.3 12.8%
4 460 28.2 20.1%
5 423 23.4 13.1%
6 387 18.8 14.9%
7 382 20.6 14.4%
8 380 22.8 30.0%
9 370 23.3 12.7%
10 359 31.9 27.0%
11 340 21.2 17.4%
12 336 21.3 12.6%
13 333 22.2 23.1%
14 311 21.1 16.4%
15 308 22.7 21.8%
16 307 22.0 18.8%
17 301 17.9 8.8%
18 293 21.5 14.4%
19 286 27.2 24.7%
20 279 26.0 15.4%
21 253 15.8 6.1%
22 253 22.9 15.7%
23 243 19.4 16.9%
24 238 25.4 21.1%
25 232 25.8 32.4%

Source: Milliman analysis of 9/22/2022-9/22/2023 Medicare 100% RIF data. Analysis based on 627,095 SNF stays among 5,545 total SNFs.

Conclusion

This analysis suggests that significant variation exists in SNF performance for ALOS and readmission rates both among SNFs within a geographic region and among geographic regions. The variation is not explained by differences in patient case mix, which indicates the variation may be driven by differences in practice patterns among SNFs. In particular, the practice of evaluating the need for continued stay on a weekly basis instead of more frequently and the practice by some SNFs to send patients back to the hospital when their condition mildly worsens instead of stepping up care in the nursing home, can contribute to longer lengths of stay and higher readmission rates back to hospital. These findings highlight the opportunity for SNFs to more efficiently manage LOS and readmissions for Medicare beneficiaries during SNF stays, which can meaningfully reduce Medicare costs.

For MA plans, Medicare ACOs and Medicare bundled payment participants, it is essential to evaluate and profile performance for SNFs within their network and develop strategies for incentivizing more efficient performance. Contracting and referral strategies should incentivize SNFs with more efficient performance. For more information about evaluating the efficiency of SNF providers, contact your Milliman consultant.

As the overall rate of SNF admissions declines, and a larger proportion of Medicare beneficiaries are discharged from acute care hospital stays directly to the community, the remaining SNF stays may become more complex and may require longer lengths of stay.


1 MedPAC (March 2024). Report to the Congress: Medicare Payment Policy, Chapter 6: Skilled Nursing Facility Services. Retrieved September 5, 2024, from https://www.medpac.gov/wp-content/uploads/2024/03/Mar24_Ch6_MedPAC_Report_To_Congress_SEC.pdf.

2 Medicare.gov. Skilled Nursing Facility Care. Retrieved September 5, 2024, from https://www.medicare.gov/coverage/skilled-nursing-facility-snf-care.


Explore more tags from this article

About the Author(s)

We’re here to help