Trends In Hospitals And Skilled Nursing Facilities Sharing Medical Providers, 2008–16

Abstract

Hospitals and skilled nursing facilities (SNFs) face increasing pressure to improve care coordination and reduce unnecessary readmissions. One strategy to accomplish this is to share physicians and advanced practice clinicians, so that the same providers see patients in both settings. Using 2008–16 Medicare claims, we found that as SNFs moved increasingly toward using SNF specialists, there was a steady decline in the number of facilities sharing medical providers and in the proportion of SNF primary care delivered by provider practices with both hospital and SNF clinicians (hospital-SNF practices). In SNF fixed effects analyses, we found that SNFs that increased primary care visits by hospital-SNF practices had slightly fewer readmissions, shorter lengths-of-stay, and increased successful community discharges. These findings suggest that SNFs that share medical providers with hospitals may see some benefit from that linkage, although the magnitude of the benefit may be small.

Hospitals and skilled nursing facilities (SNFs) are increasingly held accountable for the quality of care provided in SNFs, particularly with respect to hospital readmissions. In 2016 roughly one in five hospitalized Medicare beneficiaries were discharged to a SNF, and up to 11 percent of those were readmitted with potentially avoidable conditions.¹ Readmission penalties and alternative payment models instituted under the Affordable Care Act, as well as the addition of readmission metrics to Hospital Compare and Nursing Home Compare, have created incentives for hospitals and SNFs to improve care coordination between settings. Despite this, there is still limited understanding regarding the kinds of interorganizational linkages that best facilitate care coordination.

Evidence to date suggests that greater hospital-SNF integration does achieve benefits. Hospitals that own SNFs or have a formal network of preferred SNFs have fewer readmissions.^2–5 Higher discharge volume from a hospital to a given SNF has also been linked to fewer readmissions,^6,7 although findings are mixed.² Both hospitals and SNFs have incentives to integrate: hospitals to exert influence over SNF costs and quality⁸ and SNFs to maintain a constant flow of referrals to justify the substantial capital and labor costs associated with providing postacute care.^9,10

Although the literature does suggest that hospital ownership of SNFs or a higher volume of exchanges between a hospital and a SNF is beneficial, it remains unclear how such arrangements affect clinical practice. The literature conjectures that shared staffing and capital (for example, the use of information technology) are the main inputs of linkage. Qualitative research suggests that many hospitals and SNFs share physicians and advanced practice clinicians (nurse practitioners and physician assistants) across settings as a way to improve coor dination,^3,8,11 and a few specific care models have reduced readmissions.^12,13 However, there is a lack of quantitative evidence to support whether this is a widespread trend.

Historically, the majority of nursing home medical providers have been community-based primary care physicians who see patients across office, hospital, and nursing home settings but spend limited time physically present in the nursing home.^14,15 During the past decade, however, there has been a steady increase in the number of physicians and, in particular, advanced practice clinicians who specialize in nursing home care.^16,17 These providers, known as SNFists, bill at least 90 percent of their visits in nursing homes^16,17 and see few, if any, patients in other settings. Whether this growth in SNFists has corresponded with a decline in providers who split their clinical time between SNFs and hospitals is unknown.

We used 2008–16 administrative claims data to examine longitudinal trends in SNF primary care visits made by SNFists and other clinicians in two types of provider practices: those that see both hospital and SNF patients (hospital-SNF practices) and those that see only SNF patients (SNF practices).We then tested whether changes in the proportion of visits within a SNF made by these different types of providers was associated with reductions in SNF rates of thirty-day hospital readmission, successful discharge to community, and SNF length-of-stay.

Study Data And Methods

Data Sources

We created our study cohort by linking 2008–16 data from the following sources: the Medicare Master Beneficiary Summary File, Medicare Part A inpatient and SNF claims, Medicare Part B Carrier Summary Data File claims, and the Minimum Data Set. The Carrier file includes all claims for a 20 percent random sample of Medicare beneficiaries, whereas the other files include data on all beneficiaries. To obtain SNF characteristics and our outcome measures, we linked these data to LTC Focus, a publicly available database from Brown University, derived from Certification and Survey Provider Enhanced Reports, Minimum Data Set, Area Resource File, and aggregated Medicare claims.¹⁸ We also merged in the Medicare Shared Savings Program file and the Center for Medicare and Medicaid Innovation model participant roster¹⁹ to identify SNFs that were part of accountable care organizations or the Bundled Payments for Care Improvement initiative Model 3: Retrospective Post Acute Care Only.

Provider Practices

We identified provider practices using the Taxpayer Identification Number submitted on Carrier file claims. We limited our analysis to claims by primary care physician specialties (general practice, family practice, internal medicine, osteopathic medicine, geriatrics) and advanced practice clinicians (nurse practitioners, physician assistants) for evaluation and management visits in offices, hospitals, and nursing homes. We identified individual clinicians using the National Provider Identifier and, as done in prior work,^16,17 classified them as SNFists if at least 90 percent of their total billed visits were in nursing homes. This definition is comparable to that of hospitalists.^20,21 Clinicians were linked to a practice if they submitted at least one claim under the practice’s Taxpayer Identification Number within a six-month window.

Hospital–Skilled Nursing Facility Linkage

Carrier file claims identify the type of setting in which an encounter takes place, but not the specific health care organization. To obtain this information, we merged the Carrier file with data from Part A claims and the Minimum Data Set, using the Residential History File methodology,²² an algorithm that concatenates Medicare claims and Minimum Data Set assessments to produce a daily, per person chronological history of health care use across settings. We used this to identify all Medicare fee-for-service beneficiaries in the 20 percent Carrier file who were discharged directly from an acute general hospital to a SNF between January 1, 2008, and December 31, 2016, excluding beneficiaries who were nursing home residents at the time of hospitalization. We then used the Medicare Beneficiary Identifier and dates of service to identify provider visits that occurred during specific hospital and SNF episodes for those beneficiaries.

Classifying Primary Care Visits

We classified physicians and advanced practice clinicians seeing SNF patients according to whether or not the individual clinician was a SNFist and whether the clinician’s practice saw both hospital and SNF patients (hospital-SNF practice) or SNF patients only (SNF practice). All practices could also see office patients, but we defined the two practice types based solely on hospital and SNF patients. This resulted in four types of providers for analysis. Among hospital-SNF practices there were non-SNFists (clinicians who see hospital, SNF, and office patients) and SNFists (clinicians employed by the practice to follow only that practice’s SNF patients). Among SNF practices there were non-SNFists (clinicians who see SNF and office patients, but no hospital patients) and SNFists (clinicians employed by a SNF-based practice or by an office-based practice to see that practice’s SNF patients).

To identify hospital-SNF practices, we applied the following criteria to ensure that the practice had a regular presence in both hospitals and SNFs:Within six months, the practice saw at least twenty hospital and SNF patients; hospital and SNF encounters accounted for at least 10 percent of the practice’s total encounters; of those encounters, at least 10 percent were in a hospital and 10 percent in a SNF; and hospital and SNF patients accounted for at least 10 percent of the practice’s total patients. We chose the twenty-patient threshold using the 20 percent Carrier file to approximate one hundred patients in the full population, an approach adapted from others who have used partial sample files.²⁰ We applied the other criteria to reduce the likelihood of classifying a practice that only sporadically appeared in either setting as a hospital-SNF practice. We generated our counts based on a six-month time span to account for potential midyear contract changes among practices, hospitals, and SNFs.

We next classified a hospital and a SNF as sharing medical providers if, within six months, at least two patients in the hospital and at least two patients discharged from that hospital to that SNF were seen by the same practice (as identified by Taxpayer Identification Number). Because both hospitals and SNFs require medical providers to undergo a credentialing process to see patients and bill in that organization,^23,24 we made the assumption that any appearance of a clinician in that setting implied that the clinician had some formal arrangement to practice there. However, we set the threshold to two patients to allow for error. For both the hospital-SNF practice and hospital-SNF pairs, we conducted sensitivity analyses using more and less stringent criteria.

We identified a total of 8,877,094 SNF primary care visits (see online appendix exhibit 1) for 1,494,113 Medicare beneficiaries from 2008 to 2016.²⁵ This represented a mean of seventy-eight visits per SNF per year (data not shown). We calculated the percentage of visits across all claims and within each SNF completed by the four types of medical providers.

Outcomes

We used three SNF-level measures from LTCFocus as our outcomes. The thirty-day readmission rate is the percentage of total new admissions from hospitals to the SNF who have a Minimum Data Set assessment showing discharge to an acute hospital within thirty days of SNF admission. The successful discharge to community rate is the percentage of total new admissions from hospitals to the SNF who have a Minimum Data Set assessment showing discharge to the community within one hundred days of SNF admission and no subsequent Minimum Data Set assessment for any facility within thirty days of SNF discharge. Median length-of-stay is the median number of days between SNF admission and discharge for all new admissions from hospitals to the SNF. All outcome measures are unadjusted.

Statistical Analysis

Although our descriptive statistics are reported for 2008–16, we limited our regression analysis to 2011–16 because our outcome measures are all derived from Minimum Data Set version 3.0, which was implemented in October 2010. We used linear regression models with SNF and year fixed effects to examine the effects of a 10 percent increase in a SNF’s primary care visits by the different provider types (using non-SNFists in SNF practices as the base category) on each of our three outcomes. We controlled for time-variant SNF characteristics including Medicaid, Medicare, and Medicare Advantage censuses; admissions per bed; number of primary care claims per SNF; and staffing measures for nursing and rehab staff. Analyses were conducted using SAS, version 9.4, and Stata, version 15.1.

Limitations

Some limitations of our study should be noted. First, using claims data, we were unable to ascertain information about a practice’s ownership or organizational structure. That is, we knew the settings in which the practice’s providers saw patients, but nothing about other administrative, financial, or technical mechanisms by which these practices may have been aligned with those health care organizations. Second, as is the case with all observational analyses, there could be unobserved correlated factors influencing our results.

Third, our sample of SNFs was derived from claims for a 20 percent random sample of Medicare beneficiaries rather than for all beneficiaries. Although only a 20 percent file was used, almost all hospitals, SNFs, and provider practices that care for sizable Medicare populations are represented. Of the roughly 14,500 SNFs in the US that accept postacute Medicare patients, we captured claims for a mean of 14,210 SNFs per year, although the number of claims per SNF varied, and not all SNFs were represented in all years. To account for this in our regression models, we controlled for the number of claims per SNF.We also conducted two sensitivity analyses, first limiting to SNFs with ten or more claims per year (n = 12,404) and then using only SNFs with data for all years (n = 10,432) (data not shown). Results were similar.

Finally, using the 20 percent file, we may have misclassified some clinicians as SNFists by not having their full patient panels, although we used similar methods as others who have defined SNFists and hospitalists with partial sample files.^17,20 We also may have misclassified some advanced practice clinicians as SNFists as a result of “incident to” billing in office settings, in which an advanced practice clinician bills under a collaborating physician’s National Provider Identifier (this practice is not allowed in hospitals or SNFs).²⁶ However, we expect that this potential for misclassification would only affect a small number of clinicians right at the 90 percent threshold—roughly 2 percent of all clinicians.

Study Results

The total number of SNF primary care visits grew 40 percent from 850,285 in 2008 to 1,189,553 in 2016, mirroring the growth rate for evaluation and management visits in the Medicare population overall.¹ Within this time frame, the number of visits by SNFists in hospital-SNF practices and SNF practices grew by 92 percent and 113 percent, respectively. Visits by non-SNFists in SNF practices grew 36 percent, whereas visits by non-SNFists in hospital-SNF practices decreased 22 percent (appendix exhibit 1).²⁵

Non-SNFists in SNF practices were the dominant providers across the study period, delivering 38 percent of visits in 2008 and 37 percent of visits in 2016, with a slight downturn after 2014. Non-SNFists in hospital-SNF practices were the second most common provider in 2008–09, but declined significantly from 31 percent of visits in 2008 to 18 percent in 2016. This coincided with a steady increase in SNFists in SNF practices, who provided 24 percent of visits in 2008 and 36 percent in 2016 and overtook non-SNFists in hospital-SNF practices as the second most common provider in 2010. SNFists in hospital-SNF practices were the least common providers across years but grew from 7 percent of visits in 2008 to 9 percent in 2016 (exhibit 1). Advanced practice clinicians represented 62 percent of all SNFists and billed 55 percent of SNFist encounters in 2008 and represented 68 percent of all SNFists and billed 64 percent of SNFist encounters in 2016 (data not shown).

Exhibit 1. Percent of skilled nursing facility (SNF) primary care visits by hospital-SNF practices and by SNF practices, 2008–16.

SOURCE Authors’ analysis of 2008–16 Medicare claims linked to the Minimum Data Set. NOTE SNFist is skilled nursing facility specialist.

Exhibit 2 shows the number of hospitals and SNFs that shared medical providers between 2008 and 2016, as well as provider practices that saw patients in both settings. The number of hospitals decreased from 2,345 in 2008 to 1,658 in 2016, whereas the number of partnering SNFs per hospital increased from a mean of 3.74 in 2008 to 3.89 in 2016, demonstrating a slight trend toward consolidation. Similarly, the number of SNFs sharing providers with hospitals decreased from 6,966 in 2008 to 4,773 in 2016, whereas the mean number of partnering hospitals per SNF increased modestly from 1.26 in 2008 to 1.35 in 2016. The number of provider practices seeing patients in both settings decreased from 5,218 in 2008 to 2,615 in 2016, whereas the mean number of hospitals per practice increased from 1.22 to 1.37 and the number of SNFs per practice increased from 2.23 to 2.77 across those same years. The increased number of hospitals and SNFs per practice may be due in part to a shift toward larger practices (appendix exhibit 2).²⁵ In 2008, 75 percent of shared hospital-SNF practices were solo or small group practices (two to six providers), and only 25 percent were large group practices (seven or more providers). By 2016, 64 percent of practices were solo or small group practices and 36 percent were large group practices.

Exhibit 2.

Numbers of hospitals, skilled nursing facilities (SNFs), and provider practices sharing medical providers, 2008–16

	2008	2009	2010	2011	2012	2013	2014	2015	2016
Number of hospitals	2,345	2,219	2,079	2,037	1,930	1,876	1,819	1,794	1,658
Mean number of SNFs with which hospitals share medical providers	3.74	3.72	3.58	3.60	3.60	3.69	3.75	3.95	3.89
Number of SNFs	6,966	6,512	5,857	5,722	5,376	5,290	5,216	5,264	4,773
Mean number of hospitals with which SNFs share medical providers	1.26	1.27	1.27	1.28	1.29	1.31	1.31	1.34	1.35
Number of provider practices	5,218	4,756	4,133	3,970	3,522	3,301	3,037	3,035	2,615
Mean number of hospitals per practice	1.22	1.23	1.24	1.25	1.27	1.30	1.32	1.34	1.37
Mean number of SNFs per practice	2.23	2.26	2.24	2.27	2.36	2.47	2.59	2.70	2.77

SOURCE Authors’ analysis of 2008–16 Medicare claims linked to the Minimum Data Set.

SNFs that share medical providers with at least one hospital differ in many ways from SNFs that do not share medical providers. Exhibit 3 illustrates these differences, using 2016 data. SNFs that share medical providers with hospitals are, on average, larger, have higher occupancy, are more likely located in urban settings, and operate in more competitive counties compared with SNFs without shared providers. They also, on average, have more Medicare residents, fewer Medicaid residents, and higher nurse and rehab therapist staffing levels, which suggests that these SNFs generally provide more postacute care than SNFs that do not share medical providers with hospitals. SNFs with more formal hospital ties, including those that are hospital based, part of an accountable care organization, or a participant in the Bundled Payments for Care Improvement initiative Model 3: Retrospective Post Acute Care Only, are more likely to share medical providers with hospitals, although the number of SNFs in these types of arrangements is relatively small. SNFs with shared providers account for roughly 30 percent of SNFs overall and include 41 percent of SNFs that are in accountable care organizations, 46 percent of Bundled Payments for Care Improvement Model 3 SNF participants, and 33 percent of hospital-based SNFs.

Exhibit 3.

Characteristics of skilled nursing facilities (SNFs) that share and do not share medical providers with at least one hospital, 2016

Characteristics	No shared providers (n = 10,768)	Shared providers (n = 4,557)
Organizational characteristics
Total beds, mean	100.3	120.2****
Beds occupied, %, mean	80.0	81.8****
For-profit ownership, %	69.7	68.2*
Owned by a chain, %	56.9	59.0**
Urban location, %	63.9	83.7****
Herfindahl-Hirschman Index for nursing home beds in county	0.34	0.23****
Hospital-based, %	4.9	5.7**
BPCI Model 3 participant, %	1.5	2.9****
Part of an accountable care organization, %	2.5	4.1****
Payer mix, %
Residents with Medicaid as primary payer, mean	61.8	50.0****
Residents with Medicare as primary payer, mean	12.0	21.5****
Staffing, FTEs per 100 beds, mean
Registered nurses	8.0	10.3****
Licensed practical nurses	13.7	14.7****
Certified nursing assistants	37.5	39.3****
Physical therapists	1.4	2.5****
Occupational therapists	1.3	2.3****
Speech language pathologists	0.84	1.12****
Physicians	0.38	0.43**
Nurse practitioners and physician assistants	0.3	0.39*
Any nurse practitioner or physician assistant on site, %	50.5	60.7****
Specialty units, %
Alzheimer unit	14.8	16.5***
Special rehab unit	2.0	2.9***
Ventilator unit	2.2	2.6*

SOURCE Authors’ analysis of 2016 Medicare claims, Minimum Data Set, LTCFocus, Medicare Shared Savings Program file, and Center for Medicare and Medicaid Innovation model participant roster. NOTES Two hundred sixteen SNFs with shared providers did not match to the Certification and Survey Provider Enhanced Reports and are thus excluded from the exhibit. Differences in SNF characteristics tested using analysis-of-variance tests for continuous variables and Pearson chi-square tests for categorical variables. Percentages shown are percentages of the subsamples shown in the column heading. BPCI is Bundled Payments for Care Improvement initiative. FTE is full-time equivalent.

^*p < 0:10

^**p < 0:05

^***p < 0:01

^****p < 0:001

Appendix exhibit 3 shows the mean annual distribution in the proportion of primary care visits per SNFby the four provider types (SNFists and non-SNFists in hospital-SNF practices and SNFists and non-SNFists in SNF practices).²⁵ Exhibit 4 shows the estimated relationships between changes in those proportions and SNF-level measures of thirty-day readmissions, successful discharge to community, and median length-of-stay.We used non-SNFists in SNF practices as the base category. In linear regression models controlling for SNF and year fixed effects as well as time-variant SNF characteristics, we found that SNFs that increased the share of visits by hospital-SNF practices had small but statistically significant beneficial effects across outcomes.

Exhibit 4.

Estimated effects of a 10 percent increase in a skilled nursing facility’s (SNF’s) primary care visits by SNFists and non-SNFists in different practice types on outcomes, 2011–16

	Outcomes
	30-day readmissions, mean		Successful discharge to community, mean		Median length-of-stay
Explanatory variables	Amount/coefficient	95% CI	Amount/coefficient	95% CI	Amount/coefficient	95% CI
Sample mean or median	18.0%	—a	54.2%	—a	36.0 days	—a
Percent of visits by hospital-SNF practices
SNFists	−0.03**	(−0.06, −0.003)	0.07**	(0.01, 0.13)	−0.08	(−0.17, 0.001)
Non-SNFists	−0.02	(−0.04, 0.002)	0.12****	(0.08, 0.15)	−0.14****	(−0.20, −0.09)
Percent of visits by SNF practices
SNFists	−0.01	(−0.03, 0.003)	−0.02	(−0.06, 0.01)	0.05	(−0.01, 0.11)
Observations	77,283	—a	77,094	—a	77,113	—a
Number of SNFs	14,680	—a	14,631	—a	14,631	—a
R2	0.65	—a	0.81	—a	0.69	—a

SOURCE Authors’ analysis of 2011–16 Medicare claims linked to the Minimum Data Set and LTCFocus. NOTES Coefficients represent the estimated effects of a 10 percent increase in the share of a SNF’s primary care visits by SNFists and non-SNFists within hospital-SNF practices and SNFists within SNF practices compared with non-SNFists within SNF practices in linear regression models with SNF and year fixed effects. Control variables include SNF and year fixed effects, Medicaid census, Medicare census, Medicare Advantage census, admissions per bed, number of primary care visits per SNF, and staffing full-time equivalents per one hundred residents for registered nurses, licensed practical nurses, certified nursing assistants, and licensed rehab therapists. Models were estimated using the areg command in Stata. CI is confidence interval.

^aNot applicable.

^**p < 0:05

^****p < 0:001

Holding visits by other provider types constant, a 10 percent increase in share of visits by SNFists in hospital-SNF practices, relative to non-SNFists in SNF practices, decreased thirty-day readmissions by 0.03 percent and increased successful discharges to community by 0.07 percent. A 10 percent increase in the share of visits by non-SNFists in hospital-SNF practices increased successful discharges to community by 0.12 percent and decreased median length-of-stay by 0.14 days. Differences in the share of visits by SNFists in SNF practices versus non-SNFists in SNF practices were not statistically significant for any of the outcomes. A full specification of our regression models is available in appendix exhibit 4.²⁵

To test for a potential dose response with regard to the degree of hospital-SNF integration, we dichotomized the four provider types to represent whether a SNF was above versus at or below the mean in the proportion of visits by each type. Results were similar, although effect sizes were slightly larger (appendix exhibit 5).²⁵

Discussion

From 2008 to 2016 the rise in clinical SNF specialists, or SNFists, coincided with a steady decline in the proportion of primary care delivered in SNFs by clinicians who split time between hospitals and SNFs. There was also a significant decline in the number of hospitals and SNFs that share medical providers, as well as the number of provider practices that follow patients in both settings.We found particularly sharp declines in the number of solo and small group practice providers who split their time between hospital and SNF settings, reflective of broader trends in practice consolidation.²⁷

Even within provider practices that continue to follow both hospital and SNF patients, those practices are increasingly using SNFists to see their SNF patients. These SNFists, similar to SNFists in SNF practices, are predominantly advanced practice clinicians. These clinicians have taken on an increasingly prominent role in SNFs as a result of growing shortages of geriatricians and primary care physicians—particularly those with the interest and skill to see nursing home patients—as well as increasing care complexity in this setting, requiring closer medical oversight.^15,28 Non-SNFists in SNF practices (clinicians who see both office and SNF patients) remained the dominant providers across years, although in 2016 they provided only slightly more visits than did SNFists in SNF practices. Overall, these findings suggest that the historically dominant model of SNF medical care being delivered by community primary care providers who split clinical time across settings and see SNF patients as one aspect of a larger practice is being replaced by one in which SNFists are playing a larger role.

Although the overall trend appears to show a decline in hospital-SNF linkages through shared medical providers, there is a subset of SNFs that are integrated with hospitals in this way. SNFs that share medical providers with hospitals appear to be more specialized in postacute care. They typically are larger, urban SNFs that are more nursing and rehab intensive; accept more short-stay Medicare Part A patients; and are more likely to have formal hospital ties through shared ownership, accountable care organizations, or bundled payment models. We could not conclude from our data whether SNFs and hospitals with these more formal partnerships entered those arrangements in part because of preexisting relationships that involved shared medical providers, or whether the sharing of providers came about as a result of those partnerships.

It also may be that broader trends in hospital-provider practice consolidation are influencing the changes in hospital-SNF relations we see over time. That is, practices that remain consolidated with hospitals in the later years might be of better quality and are most likely larger. As such, they may be better positioned to align themselves with SNFs to facilitate hospital-SNF care coordination. It is beyond the scope of this article to examine the contribution of hospital and practice quality on decisions surrounding consolidation, but the effects of such selection on our estimated relationships would likely be minimized by the presence of SNF fixed effects in our models.

We found that SNFs that increased primary care visits by hospital-SNF practices had slightly better outcomes (lower readmission rates, shorter lengths-of-stay, and more successful discharges to community) than those that did not. These differences were statistically significant, but the effect sizes were very small and might not translate to clinical significance. Interestingly, we found significant effects for both SNFists and non-SNFists within hospital-SNF practices, but not for SNFists within SNF practices. We hypothesize that SNFists employed within hospital-SNF practices may have the advantage of being able to collaborate with other providers in their practices who have hospital privileges to facilitate care coordination.

Potential advantages of SNF medical providers having hospital privileges are that these clinicians can access the hospital’s medical records and consult with hospital staff to obtain more detailed information on patients transitioning from the hospital to a SNF than what SNF medical providers without privileges typically have to rely on in standard hospital discharge summaries.¹² Likewise, providers with both hospital and SNF privileges may have more control over the care SNF patients receive when hospitalized. For example, they may be able to intervene to prevent emergency department visits for minor conditions from becoming unnecessary admissions. They may also be able to assist other SNF colleagues with care coordination by serving as the hospital attending of record when those other providers’ patients are hospitalized.

We did not find significant differences in SNF outcomes based on the proportion of visits by SNF practice SNFists versus non-SNFists. A recent study by Kira Ryskina and colleagues found that patients of SNFists had lower rehospitalization rates and were more likely to be successfully discharged to the community compared with patients of non-SNFists.²⁹ There are two important differences between this study and ours. First, we identified four types of providers instead of two, and we introduced the added dimension of the hospital-SNF practice. Second, Ryskina and colleagues used patient-level measures to examine the effects of having a specific provider type on patient outcomes. We instead used the proportion of visits by different provider types as a SNF-level measure of hospital linkage and examined how this related to SNF-level outcomes. It may be the case that although individual patients benefit from being cared for by SNFists, there is some benefit at the facility level for having that added level of hospital-SNF linkage.

We were unable to determine, using administrative claims and assessment data, whether other aspects of hospital-SNF linkages were in place to facilitate care coordination. An example of this would be a shared electronic medical record, or at least a hospital electronic medical record to which SNF providers have access. Hospitalized patients are often discharged to SNFs with discharge summaries containing incomplete or missing information about consults, diagnostic studies, and other events that occurred during the hospitalization.³⁰ Having electronic medical record access enables SNF providers to access such information without having to rely on time-consuming follow-up medical record requests. Another strategy SNFs may use to facilitate medical care coordination with a hospital is negotiating with a designated hospitalist or inpatient service to be the attending physician or service of record for patients sent from the SNF to that hospital. Then, even if the SNF clinicians lack privileges for that hospital, they still have consistent relationships with inpatient providers to facilitate communication.

Conclusion

Hospitals and SNFs face increasing pressure from public reporting, alternative payment models, and readmission penalty programs to reduce hospital readmissions and improve transitions between settings. One avenue by which hospitals and SNFs can improve care coordination is through the sharing of physicians and advanced practice clinicians across settings. Our findings suggest that this may be beneficial for SNFs and that a subset of SNFs that appear to be more specialized in postacute care do engage in this practice. Still, the number of hospitals and SNFs sharing medical providers is declining as both hospitals and SNFs increasingly rely on hospitalists and SNFists, respectively, leaving considerable opportunity to explore other mechanisms to improve care coordination between settings.