a b s t r a c t

Introduction: We examined trends in the use of observation services and the relationship between index service type (observation services, emergency department [ED] visits, inpatient stays) and both clinical outcomes and Medicare payments.

Methods: We created a yearly cohort panel of Medicare beneficiaries with chest pain. We evaluate the relation- ships between index service type and 30-day clinical outcomes using a multinomial logit model and between index service type and Medicare payments using Generalized linear models.

Results: In 2009, 24% of patients with chest pain received observation services; this rose to 29% in 2011. Converse- ly, 20% were treated as hospital inpatients in 2009; this fell to 16% in 2011. In the adjusted analysis, the risk of 30- day return to the hospital was 7% less (95% confidence interval, 5%-8%) for those receiving observation services as compared with inpatients. Average Medicare payments ranged from $3032 for beneficiaries initially treated in the ED to $3885 for those initially treated in observation to $6545 for those initially treated as inpatients.

Discussion: Patients treated in observation are less likely than those treated in the ED or as inpatients to have an adverse event within 30 days. Adjusted Medicare payments, including the index stay and the subsequent 30 days, were substantially less for those treated in observation as compared with those treated as inpatients, but more than for those treated and released from the ED. Higher rates of observation service use do not appear to be neg- atively affecting patient outcomes and may lower costs relative to inpatient treatment.

Introduction

Since 2007, there has been consistent growth in observation services in US hospitals [1]. Current efforts to reduce health care costs have focused attention on observation services as a replacement for short- term hospitalizations in the context of reduced payment policies for excess readmissions and funding to test models for improving care tran- sitions. In addition, increased use of observation services may also be re- lated to the growth of Observation Units [2-4], the desire to reduce readmissions and short-stay admissions, and the risk of potential audits. Historically, the literature has focused on the growing use of obser- vation units for conditions such as chest pain [5,6], but recently, the focus has shifted to the appropriate use of observation [7], the impact of longer lengths of stay [8], descriptions of observation care, and out- comes for specific diagnoses [9,4]. An Office of the Inspector General re- port from 2013 describes observation services and contrasts them with

* Corresponding author at: Research and Rapid-Cycle Evaluation Group, Center for Medicare and Medicaid Innovation, Mail Stop WB-06-05, 7500 Security Boulevard, Baltimore, MD 21244.

E-mail address: [email protected] (S.G. Cafardi).

short inpatient stays. Based on their memorandum, the use of observa- tion services was shown to reduce the average cost of a hospital visit for both beneficiaries and Medicare [10].

In this article, we build upon the Office of the Inspector General work

that focuses on the initial treatment setting and Associated costs. We first describe the shift in patterns of care (whether a patient is discharged from the emergency department [ED], receives observation services, or is admitted) for beneficiaries with chest pain between 2009 and 2011. Then, we examine the outcome and cost implications of this shifting by comparing 30-day return to hospital (readmissions, observation services, ED visit) and death and Costs of care for the index event and a 30-day follow-up period, respectively.

Methods

Participants

We performed a retrospective cohort study of chest pain visits using a 20% nationally Representative sample of fee-for-service (FFS) Medi- care beneficiaries for calendar years 2009 to 2011. The yearly cohort

http://dx.doi.org/10.1016/j.ajem.2015.08.049 0735-6757/

S.G. Cafardi et al. / American Journal of Emergency Medicine 34 (2016) 16-19 17

panels were constructed by identifying patients discharged alive with chest pain as a primary or secondary diagnosis on a claim submitted to Medicare for payment from a Prospective Payment System hospital. We excluded those who were found to have a myocardial infarction or other serious pulmonary or cardiovascular condition. Beneficiaries were eligible for inclusion if they were continuously enrolled in FFS Medicare for all alive and eligible months of the cohort year. Individuals were categorized by the level of care billed by the facility for the initial chest pain visit (ED, observation, or inpatient). We assessed the final level of care assigned to the patient and did not account for scenarios where the patient was moved from observation to inpatient setting prior to discharge as such events cannot be reliably ascertained using Medicare claims. The final analytic cohort included 207478 unique ben- eficiaries in 2009, 216301 in 2010, and 216692 in 2011. Each annual sample was constructed independently, and the First visit of a beneficia- ry to the hospital within a year was used. Observation services were coded if they lasted at least 8 hours-the minimum Medicare threshold for payment.

Predictor variables

Beneficiary age, sex, race, concurrent Medicaid status (ie, being dual- ly eligible for Medicare and Medicaid), and presence of end-stage renal disease were obtained for the 3 groups using the Center for Medicare & Medicaid Services chronic conditions Warehouse beneficia- ry summary file. Diagnoses from the Center for Medicare & Medicaid Services Chronic Conditions Warehouse including hierarchical condi- tion category (HCC) risk scores were used to control for comorbidities associated with health care costs [11]. We also included hospital charac- teristics from American Hospital Association data.

Outcome variables

We analyzed 30-day outcomes for beneficiaries in our sample who were discharged alive between January 1 and December 1 of each year studied. Specifically, we used return to the hospital (including readmissions, observation services, and ED visits) and all-cause mortal- ity as outcomes. We examined cost implications using total Medicare Part A and Part B payments for the index visit and the 30 days after the index discharge. Nominal payments were adjusted to reflect 2009 dollars using the medical care component of the consumer price index.

Statistical analysis

For each year, we calculated descriptive statistics and standard com- parative statistics (eg, t tests and ?² tests) to evaluate differences in ben- eficiary characteristics, settings, and outcomes. We fit a multinomial logit model of the determinants of inpatient stays and observation ser- vices to describe changes in treatment setting over time after adjusting for beneficiary and hospital characteristics. We modeled the associa- tions between treatment settings and 30-day revisit and mortality rates using a multinomial logit model. We used a generalized linear model (with log link and gamma distribution) to model the association between treatment settings and total Medicare payments. Cluster- adjusted standard errors are reported due to some overlap of beneficia- ries across years. All analyses were performed using SAS version 9.1 (SAS Institute, Inc, Cary, NC) and Stata statistical software version 12.1 (StataCorp LP, College Station, TX).

Results

Clinical and demographic characteristics of the samples remained relatively constant over time (Table 1). There was a steady increase in the proportion of beneficiaries with chest pain who were treated with observation services, a steady decrease in the proportion of those treat- ed as inpatients, and the proportion treated in the ED remained

Table 1

Characteristics of FFS Medicare beneficiaries with chest pain, 2009-2011

	2009	2010	2011
Total, n	207478	216301	216692
Type, n (%)
ED	117212 (56)	121599 (56)	121236 (56)
OBS	48820 (24)	55420 (26)	61914 (29)
INPT	41446 (20)	39282 (18)	33542 (15)
Age (y), n (%) b65	56942 (27)	60708 (28)	62463 (29)
65-74	64626 (31)	67152 (31)	67320 (31)
75-84	56478 (27)	57655 (27)	56402 (26)
>= 85	29432 (14)	30786 (14)	30507 (14)
Sex, n (%)
Male	124673 (60)	128911 (60)	129591 (60)
Female	82805 (40)	87390 (40)	87101 (40)
Race, n (%)
White	168123 (81)	174515 (81)	173528 (80)
Black	29057 (14)	30893 (14)	31957 (15)
Other	10298 (5)	10893 (5)	11207 (5)
Concurrent Medicaid (ie, dual),
n (%)
Nondual	143787 (69)	147965 (68)	146652 (68)
Dual	63691 (31)	68336 (32)	70040 (32)
ESRD, n (%)
No ESRD	201395 (97)	209818 (97)	210105 (97)
ESRD	6083 (3)	6483 (3)	6587 (3)
HCC score, mean (SD)	1.6 (1.3)	1.6 (1.3)	1.6 (1.4)

Characteristics are shown by year; dual status varied by 3% across the years in the study period. All other characteristics shown varied between 0% and 2% across the years in the study period. Abbreviations: INPT = inpatient admission; OBS = observation services; ED = Emergency Department.

relatively stable. In 2009, 24% were treated in observation as compared with 29% in 2011. In 2009, 20% of beneficiaries were treated as inpa- tients as compared with 16% in 2011. Fig. 1 depicts the trends in the care setting of index visits adjusted for differences in patient and hospi- tal characteristics. There was little change in ED visits, whereas inpatient stays trended downward and observation encounters steadily rose.

Table 2 shows the results of the associations between treatment set- tings and 30-day clinical outcomes from a multinomial logit regression and between treatment settings and Medicare payments for the index service and the 30 days after using a generalized linear model. Each model controlled for patient and hospital characteristics and year of ser- vice. An alternative view of these results is shown in Fig. 2. After adjusting for Demographic and clinical factors, those who received ob- servation services were 29% less likely to return to the hospital as com- pared with those who received ED care only. Those who received inpatient services were 24% less likely to return to the hospital as com- pared with those who received ED care only. From the underlying pa- rameter estimates, we also calculate that patients who received observation services were 7% less likely to return to the hospital within 30 days (95% confidence interval [CI], 5%-8%) as compared with those treated as inpatients. After adjusting for demographic and clinical fac- tors, those who received observation services were 66% less likely to die as compared with those who received ED care only. Those who re- ceived inpatient services were 36% less likely to die as compared with those who received ED care only. Patients who received observation services were 47% less likely to die within 30 days (95% CI, 5%-8%) as compared with those treated as inpatients. Medicare payments, includ- ing the index service and for the following 30 days, were 28% higher for patients who received observation services as compared with those with an ED visit only. Payments were 116% higher for patients who had an inpatient stay as compared with those were treated in the ED only. We also calculate that Medicare payments were 41% lower for pa- tients who were treated in the observation setting as compared with those who received inpatient care.

Fig. 2 displays the adjusted 30-day rates based on the multinomial

regression model described above for the following outcomes-none, return to the hospital, and death-and adjusted Medicare payments by

18 S.G. Cafardi et al. / American Journal of Emergency Medicine 34 (2016) 16-19

Emergency department

probability

.5

.6

Inpatient stay

probability

.2

.3

Observation services

probability

.2

.3

2009 2010 2011

.3

.4

0

.1

0

.1

year

2009 2010 2011

year

2009 2010 2011

year

Fig. 1. Trends in ED, inpatient and observation services, and Medicare payments for index stay services adjusted for patient and hospital characteristics.^a. ^aED visit, inpatient stay, and ob- servation services were modeled using a multinomial logit model. Medicare payment is modeled using a generalized linear model. The rates and dollars shown in the figure, by year, are adjusted for differences in ESRD designation, sex, race, dual-eligibility status, age, HCC score, indicators for the state of the hospital, and whether the hospital was in an urban or rural set- ting, its teaching status, and indicators for bed size categories.

treatment setting. Our results showed that 21% of those with an inpa- tient stay and 18% of those with an observation stay returned to the hos- pital within 30 days. Of the 25% who initially received care in the ED and returned to the hospital within 30 days, more than half of these (14%) were treated and released from the ED on their subsequent visit. Con- versely, the likelihood of not experiencing adverse 30-day outcomes ranged from 75% for those discharged from the ED to 79% for those treated as inpatients and 82% for those initially treated in observation. The rate of death within 30 days of discharge was low among all 3 groups (between 0.2% and 0.5%). Medicare payments, including the index service and following 30 days, averaged $3032 for those treated in the ED, $3885 for those treated in observation, and $6545 for those treated as inpatients.

Discussion

From 2009 to 2011, the use of observation services for Medicare ben- eficiaries with chest pain increased by 5%, whereas inpatient admissions decreased by approximately 5%. These results are consistent with trends found in other research studies [12]. There were no meaningful changes in the rates of ED visits. This indicates that observation services are serving as a substitute for some short-stay inpatient admissions for Medicare ben- eficiaries with chest pain, whereas the decision to admit the patient for further testing after ED evaluation has not changed.

In spite of fears of worse outcomes and higher overall health care costs due to observation services, we find that the use of observation services was not associated with increased rates of return to the hospital or mor- tality compared with either those treated as inpatients or treated and re- leased from the ED. On the contrary, of the 3 treatment settings, patients treated in observation settings are most likely to have no adverse events and least likely to have subsequent visits to the ED or hospital and also least likely to die. Adjusted Medicare payments for the index stay and the next 30 days were substantially less for those treated in observation as compared with those treated as inpatients, but more for those initially treated in observation than for those treated and released from the ED. Thus, higher rate of observation services use does not appear to be nega- tively affecting value-added care delivery, and the use of observation ser- vices, when appropriate, may result in cost savings This also supplements other work that assesses the mean financial liability for Medicare benefi- ciaries receiving observation services. Between 2010 and 2012, mean beneficiary liability for a beneficiary’s first observation service was $469 (SD $442) and the mean 60-day cummulative beneficiary liability for those with at least two observation services in a sixty day period was found to be $947 (SD $804); both means are lower thn the 2010 inpatient deductible of $1100 [13].

Studies in the literature have limited their focus to the fact that

readmissions are prevalent and thereby costly [14] or have modeled po- tential cost savings associated with having observation units [15] and

Table 2

30-day clinical outcomes and Medicare payment by index service type^a

Relative risk ratio

Return to the hospital Death Payment

ED Reference Reference Reference

Observation service 0.71??(0.70-0.72) 0.34??(0.30-0.39) 1.28??(1.27-1.29)

Inpatient admission 0.76??(0.74-0.77) 0.64??(0.56-0.72) 2.16??(2.14-2.18)

b65 y Reference Reference Reference

65-74 y 0.62?? (0.60-0.63) 0.85 (0.72-1.00) 1.11?? (1.08-1.14)

75-84 y 0.64?? (0.62-0.65) 1.30?? (1.11-1.52) 1.06?? (1.03-1.09)

>= 85 y 0.67?? (0.66-0.69) 2.48?? (2.12-2.91) 0.86?? (0.84-0.89)

Male Reference Reference Reference

Female 1.10?? (1.09-1.12) 1.33?? (1.22-1.45) 1.23?? (1.21-1.25)

White Reference Reference Reference

Black 1.04?? (1.02-1.06) 0.89 (0.75-1.06) 0.89?? (0.87-0.91)

Other 0.95?? (0.92-0.98) 1.20 (0.95-1.52) 0.92?? (0.89-0.96)

No ESRD Reference Reference Reference

ESRD 1.29?? (1.24-1.33) 1.66?? (1.35-2.03) 2.84?? (2.77-2.92)

No concurrent Medicaid Reference Reference Reference Concurrent Medicaid (ie, dual) 1.26?? (1.23-1.30) 0.05?? (0.03-0.08) 0.97? (0.94-1.00) HCC score 1.17?? (1.17-1.18) 1.46?? (1.42-1.50) 1.13?? (1.13-1.14)

^a A multinomial logit model was estimated for return to the hospital, death, and no adverse outcome (the base category). A generalized linear model with log link and gamma family was estimated for Medicare payments. Both models also adjusted for state of the hospital and year indicators. 95% CIs are shown in parentheses.

* The coefficient is statistically significant at P b .05.

?? The coefficient is statistically significant at P b .01.

S.G. Cafardi et al. / American Journal of Emergency Medicine 34 (2016) 16-19 19

No adverse event Return to hospital

probability

.65 .7 .75 .8 .85

.6

probability

.05 .1 .15 .2 .25

0

ED IP OBS ED IP OBS

Death Payments for index event and 30-day followup

probability

.004

$

0 2000 4000 6000

ED IP OBS ED IP OBS

0 .002

Fig. 2. Association of return to the hospital and Death rates and 30-day Medicare payment with admission type.^a. ^aED visit, inpatient stay, and observation services were modeled using a mul- tinomial logit model. Medicare payment was modeled using a generalized linear model. The rates and dollars shown in the figure are adjusted for differences in ESRD designation, sex, race, dual- eligibility status, age, HCC score, year, indicators for the state of the hospital, whether the hospital was in an urban or rural setting, its teaching status, and indicators for bed size categories.

have not assessed the actual clinical and Financial impact of observation services as we do here.

An increasing focus on observation services has emerged amidst nu- merous initiatives that have likely influenced their use and may contrib- ute to overall decreases in hospital utilization by Medicare beneficiaries [16]. Although it is beyond the scope of this analysis to assess the effects such programs have, they likely contribute to the trends we report here. For this reason, continued monitoring and evaluation of the use of ob- servation services is needed, particularly as new payment models are promulgated that may impact care patterns in the ED.

Our findings from this retrospective claims-based analysis should be viewed in the context of numerous limitations. First, we studied associ- ations between care setting and cost and clinical outcomes in a Medi- care population, but we could not determine causal relationships. We also do not examine a comprehensive set of patient outcomes. For ex- ample, we were not able to assess the impact of increasing observation status usage on functional status and patient experience of care. In addi- tion, 30-day death rates in the 3 populations were low, ranging from 0.2% to 0.5%. These may not be directly comparable across ED and the 2 other settings because patients who died during their index visit were excluded. We used beneficiary HCC risk scores, but these are not specific to the presenting encounter. We are thereby unable to ascertain presenting Severity of disease between Levels of care.

In conclusion, the use of observation services for Medicare beneficia- ries with chest pain has increased and replaced inpatient visits over the study period. This shift was temporally associated with lower Medicare payments for hospital-based treatment of chest pain with no observable adverse clinical outcomes for the beneficiary.

Acknowledgments

The authors would like to acknowledge Erick Chuang, MS, MHS, Cen- ter for Medicare, Center for Medicare & Medicaid Services, who provid- ed policy interpretation and substantial guidance for the data analyses. The opinions expressed are solely those of the authors and do not

necessarily represent the views or policy positions of the Center for Medicare & Medicaid Services.

References

1. Feng Z, Wright B, Mor V. Sharp rise in Medicare enrollees being held in hospitals for ob- servation raises concerns about causes and consequences. Health Aff 2012;31(6):1251-9.
2. Affordable Care Act of 2010. Sec. 3025. Hospital readmissions reduction program. http:// www.hhs.gov/healthcare/rights/law/patient-protection.pdf 763-776. Accessed April 7, 2015.
3. Affordable Care Act of 2010. Sec. 3026. Community-based care transitions program. http://www.hhs.gov/healthcare/rights/law/patient-protection.pdf 776-782. Accessed April 7, 2015.
4. Wiler JL, Ross MA, Ginde AA. National study of emergency department observation services. Acad Emerg Med 2011;18(9):959-65.
5. Rydman R, Roberts R, Albrecht G, Zalenski R, McDermott M. Patient satisfaction with an emergency department asthma observation unit. Acad Emerg Med 1999;6(3):178-83.
6. Baugh C, Venkatesh A, Bohan J. Emergency department observation units: a clinical and financial benefit for hospitals. Health Care Manage Rev 2011;36(1):28-37.
7. Baugh CW, Schuur JD. Observation care-high-value care or a cost-shifting loophole?

   N Engl J Med 2013;369(4):302-5.

   Hockenberry JM, Mutter R, Barrett M, Parlato J, Ross MA. Factors associated with prolonged observation services stays and the impact of long stays on patient cost. Health Serv Res 2014;49(3):893-909.

8. Stead LG, Bellolio F, Suravaram S, Brown Jr RD, Bhagra A, Gilmore RM. Evaluation of transient ischemic attack in an emergency department observation unit. Neurocrit Care 2009;10:204-8.
9. Office of Inspector General. Memorandum report: hospitals’ use of observation stays and short inpatient stays for Medicare beneficiaries. Washington, DC: U.S. Depart- ment of Health and Human Services; 2013[https://oig.hhs.gov/oei/reports/oei-02- 12-00040.pdf. Accessed April 7, 2015].
10. Pope GC, Kautter J, Ellis RP, Ash AS, Ayanian JZ, Iezzoni LI, et al. Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ Rev 2004;25(4):119-41.
11. Medicare Payment Advisory Commission. Hospital Inpatient and Outpatient services. Report to the Congress Medicare Payment Policy; 2014 [http://www.medpac.gov/ documents/reports/mar14_ch03.pdf?sfvrsn=0. Accessed April 7, 2015].
12. Kangovi S, Cafardi SG, Smith RA, Kulkarni R, Grande D. Patient financial responsibil- ity for observation care. J Hosp Med 2015. http://dx.doi.org/10.1002/jhm.2436.
13. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med 2009;360(14):1418-28.
14. Baugh CW, Venkatesh AK, Hilton JA, Samuel PA, Schuur JD, Bohan S. Making greater use of dedicated hospital observation units for many short-stay patients could save

    $3.1 billion a year. Health Aff 2012;31(10):2314-23.

    Daughtridge GW, Archibald T, Conway PH. Quality improvement of care transitions and the trend of composite hospital care. JAMA 2014;311(10):1013-4.