The effect of decreasing length of stay on discharge destination and readmission after coronary bypass operation

The effect of decreasing length of stay on discharge destination and readmission after coronary bypass operation Richard M. J. Bohmer, MBChB, MPH, John Newell, BA, and David F. Torchiana, MD, Boston and Cambridge, Mass

Background. Over the decade of the 1990s, hospital stay after operation declined in response to prospective payment and managed care. As a result, complications previously detected and treated in the hospital may have begun to occur after discharge. In addition, discharge to nursing homes and rehabilitation hospitals may have increased. To address these questions, we used a statewide database to look at the use of postacute care and the 30-day readmission and mortality after coronary bypass operation. Methods. A modification of the Commonwealth of Massachusetts Division of Health Care Finance and Policy discharge data to include a unique patient identifier allowed us to retrospectively track patient destination at discharge and study 30-day readmission to all hospitals in the state. Results. Over the 3-year period after the institution of the unique patient identifier (1993 to 1996), postoperative length of stay after coronary bypass operation decreased from 7.4 to 6 days (19%, P < .0005), but the 30-day readmission rate (17.7%) did not increase. Discharge to rehabilitation hospitals and skilled nursing facilities rose significantly (11.7% to 23.8%), especially in the Medicare population (17.2% to 38.5%). Mortality in the 30 days after discharge remained constant at 0.3%. Conclusions. A shorter postoperative length of stay did not appear to disadvantage coronary artery bypass patients by increasing their likelihood of readmission or death. Cost savings from reduced length of stay were offset by increased use of postacute services. (Surgery 2002;132:10-5.) From the Graduate School of Business Administration, Harvard University, Cambridge, and the Cardiac Computer Center and Department of Surgery, Massachusetts General Hospital, Boston, Mass

STRATEGIES TO CONTAIN health care costs have focused on decreasing hospital admissions and reducing the costs of care of those patients who are admitted. Length of stay (LOS) has been a primary target of inpatient cost reduction efforts.1 LOS reduction in cardiac operation, the highest cost diagnosis related group (DRG) cluster at many hospitals, has been achieved using care paths, early extubation and mobilization protocols, and a focus on early discharge planning and the use of alternative sites of care.2-4 From 1990 to 1999, the national average for postoperative LOS after bypass operation diminished by 31% from 9.27 to 6.39 days (median from 8 to 5).5

Accepted for publication March 29, 2002. Reprint requests: David F. Torchiana, MD, Department of Surgery, Massachusetts General Hospital, 15 Fruit St, Room EDR-109, Boston, MA 02114. Copyright 2002, Mosby, Inc. All rights reserved. 0039-6060/2002/$35.00 + 0 11/60/125358 doi:10.1067/msy.2002.125358

10 SURGERY

The effects of these cost-driven management efforts on patient outcomes has not been well studied. Of particular concern is the possibility that preventable complications previously identified and treated in the hospital may have been shifted to a time after discharge with greater harm as a result. We hypothesized that if major problems occurred after discharge with a shortened hospital stay that hospital readmission rates would increase. In this study, we made use of a recent modification of the Massachusetts discharge database to contain a unique patient identifier. This allows tracking of readmission to all hospitals in the state, and enables the question to be addressed on a large scale at the same time that LOS reduction was a focus of hospital efforts at cost containment. METHODS Data source. The data source for the study was the Uniform Hospital Discharge Data Set. We examined all hospital discharges in the Commonwealth of Massachusetts from October 1, 1993, to September 30, 1996. The data set was

Surgery Volume 132, Number 1

Bohmer, Newell, and Torchiana 11

Fig 1. Flow chart of inclusion, exclusion criteria. UHDDS, Uniform Hospital Discharge Data Set; FY, fiscal year.

compiled by the Commonwealth of Massachusetts Division of Health Care Finance and Policy from the mandatory submission by each Massachusetts hospital of a set of demographic, clinical, and financial data for each hospital admission. Beginning October 1, 1993, Massachusetts hospitals added a unique patient identification number to their submission thereby allowing the identification of patients readmitted to any hospital subsequent to an index admission. In this database, the patient’s identity is encrypted for confidentiality purposes, and individuals are identified only by their scrambled social security number. Release of the data for this study was approved by the Internal Review Board of the Massachusetts Division of Health Care Finance and Policy. Patient selection. Hospital admissions were selected if they had any of the primary or secondary procedure codes for coronary artery bypass graft (CABG) (International Classification of Diseases, 9th Revision, Clinical Modification codes 3610 to 3616 inclusive plus 3619). This was taken as the index admission. Patients were excluded if, during the index admission, there was a concurrent valve replacement or other cardiac surgical

procedure, the field coding preoperative days was empty, or the patient died. Patients with a postoperative LOS > 15 days were excluded from the evaluation of readmission on the assumption that these were complicated cases that may not have been targeted for LOS reduction. Patients were excluded if they did not have a Massachusetts zip code of residence, on the assumption that if these patients were readmitted, they might have gone to an outof-state hospital. Finally, patients operated on in September 1996 were excluded because, for these patients, readmission within 30 days may have occurred in October 1996 and thus would not be included in the data set. Data analysis. A data set of CABG patients was created in which each row was loaded with data from the index admission along with all relevant data for the chronologically first recognized readmission, if there was one. From this table, a biomedical data processing (BMDP6) data set was created, and subsequent analyses were carried out with statistical software (SPSS, Inc, Chicago, Ill). Cross tabulations were performed with program BMDP4F, and the logistic model for probability of readmission was created with program BMDPLR

12 Bohmer, Newell, and Torchiana

Surgery July 2002

Fig 2. Trend in percentage of patients discharged less than 5 days, 5 to 7 days, 8 to 10 days, 11 to 15 days, and greater than 15 days postoperatively.

Fig 3. Percentage of patients discharged home, home with services, and to postacute facility.

using forward stepping. Candidate predictors of readmission were created from the index Uniform Hospital Discharge Data Set record. A HosmerLemeshow test was used to confirm goodness-of-fit of the logistic model. Time trends were tested for linear and higher order components and confirmed with linear regression analysis (BMDP2R). Comparisons among groups were tested with the appropriate analysis of variance (BMDP7D). Tests for differences in tabular distributions were conducted with the chi-square test. Unless otherwise noted all changes are statistically significant at the P < .001 level, and all data are presented as mean ± SD unless stated differently.

2.6 days to 6.0 ± 2.3 days (P = .0005). The number of patients discharged after fewer than 5 days almost tripled (7.3% to 20.3%), and the population of patients discharged after more than 10 days fell by about the same ratio (23% to 8.8%) (Fig 2). The mean age of patients increased from 63 (± 14) years old to 66 (± 10) years old. Emergency operations increased from 35% to 43%, whereas elective operations decreased from 37% to 31% over the time interval. These numbers suggest an increase in severity of illness in this population. There was no change in either the gender distribution of the population or the percentage of patients with renal failure or diabetes. The inhospital mortality of the study population was 2.4%. This did not change over the period of the study. Discharge destination. The distribution of discharge disposition changed during the period studied. Patients discharged home without any sort of home care services decreased from 50.1% to 29.6%, whereas patients discharged home with services (a visiting nurse) or to another facility increased from 37.9% to 46.2% and 11.7% to 23.8%, respectively (Fig 3). The mean age of patients discharged home, home with services, and to a facility was 62, 65, and 73 years old, respectively. Across the time period studied, 46% of those 50 years old and younger were discharged with services or to a facility compared with 78% of 75- to 80-year-olds and 85% of patients over 80 years old. In Medicare patients, the rate of discharge to a facility more than doubled from 17.2% to 38.5%. Patients with a longer LOS were more likely to be sent to a facility, and patients with a shorter LOS were more likely to be sent home. The percentage of all patients with a postoperative LOS of fewer than 5 days dis-

RESULTS The complete discharge set contained 2,379,979 patient discharges for the period from October 1, 1993, to September 30, 1996. Of these, 19,108 had CABG listed as a primary or secondary procedure. After excluding CABGs combined with other cardiac surgical procedures (2161 patients), patients with a postoperative LOS greater than 15 days (1212 patients), patients operated on in the last month of fiscal year 1996 (September 1996, 325 patients), patients with an out-of-state zip code of residence (1318 patients), in-hospital deaths (339 patients), and an incomplete record (222 patients); 13,559 CABG patients were included in the population in which we studied readmission (Fig 1). These patients were operated on in 1 of 12 institutions in the state that performed CABG operations at that time. The general characteristics of each hospital’s patient mix is given in Table I. LOS. Postoperative LOS decreased during the time of the study. Across all institutions, the average postoperative LOS fell about 20% from 7.4 ±

Bohmer, Newell, and Torchiana 13

Surgery Volume 132, Number 1

Table I. Characteristics of the patients in the sample, index hospitals coded A to L A Mean age, y 66.2 (SD) (9.9) Male (%) 73.5 Urgent (% ) 57.0 CHF(%) 16.8 Renal failure (%) 5.4 Diabetes (%) 27.7 D/C to facility (%) 28.6 D/C home with 24.6 services (%) D/C to home (%) 46.8 Readmission (%) 15.3

B

C

D

E

F

G

H

I

65.0 66.0 65.6 65.0 64.9 66.0 63.6 (10.7) (10.2) (10.4) (10.1) (10.3) (10.3) (10.3) 72.1 75.3 73.6 69.9 70.5 70.1 76.8 61 64.7 65.4 63 72.2 68.6 53.5 13.8 22.6 12.8 18.1 10.9 19.1 22.7 2.5 4.3 2.2 4.1 4.3 5.1 3.1 27.7 25.1 26.1 30.1 27.8 35.5 26.3 15.9 25 19.7 20.8 13 22.5 11.8 52 15.2 29.1 36.3 35.9 48.5 87 32.1 18.5

59.8 20.9

51.2 15.4

42.9 20.6

51.1 17.9

29 17.7

1.2 21.4

J

65.3 (10.3) 73.9 71.2 13.9 6.2 29.2 6.6 93.4 0 16

K

L

Total

66.2 61.6 66.8 (9.4) (18.4) (10) 73.4 71.9 67.4 49.1 96.1 60.8 17 10.3 21.7 3.3 2.8 5.2 21.3 28.8 25.9 8.5 10.9 10.4 36.1 23.8 19.8 55.4 15

65.3 19.5

65.2 72.4 65 16.5 4.2 28.3

69.8 18

CHF, Congestive heart failure; D/C, discharged.

Table II. Distribution of readmission DRG clusters for all readmitted patients, those readmitted to a different hospital, and those readmitted to the same hospital DRG Cardiac CVA/TIA Respiratory Infection Other

Total sample (%)

Different hospital (%)

Same hospital (%)

P value

44.7 2.4 10.2 21.9 20.8

52.3 2.4 9.2 14.0 22.2

36.1 2.5 11.3 30.8 19.4

< .00001 ns ns < .00001 ns

CVA/TIA, Stroke or transient ischemic attack; ns, not significant.

charged to a facility was 3.9%, whereas that of patients with a postoperative LOS of 11 to 15 days was 41.8%. Readmission. The unadjusted 30-day post-CABG readmission rate (readmitted patients as a percentage of all CABG patients) was 17.7% (20.6% for Medicare patients and 14.5% for nonMedicare patients). Readmitted to the institution that performed the operation were 8.3%, and 9.4% were admitted to a different Massachusetts hospital, a ratio that was consistent over time. This phenomenon has been observed previously.7,8 Readmission rates did not change over the period of the study. Of those patients discharged to a facility, home with services, or home unassisted, the unadjusted readmission rates were 27.1%, 17.2%, and 14%, respectively. The average LOS for readmitted patients was 6.7 ± 9 days (median 5 days, interquartile range 2 to 8 days; 0.3% of patients discharged within any given quarter died within 30 days of their discharge. Mortality was constant over all quarters observed. The distribution of the DRGs of the readmitted patients is shown in Table II. We grouped the 377 readmission DRGs into 5 categories: cardiac, stroke, respiratory, infection, and other. Patients readmitted with a cardiac diagnosis (most commonly atrial

fibrillation or congestive heart failure) were more likely to be admitted to their local hospital. Patients in the infection DRG were more likely to return to the site of their operation. Independent predictors of readmission were institution index admission LOS (longer LOS was associated with an increased likelihood of readmission), discharge disposition (patients discharged home with services or to a postacute facility were more likely to be readmitted), increased age, female gender, urgency of initial CABG operation, the presence of comorbid disease (congestive heart failure, renal failure, and diabetes), and atrial fibrillation (Table III). In our logistic model at any level of severity, a shorter LOS predicted a lower probability of readmission. This effect was independent of time. For patients with the same severity of illness, a patient with a shorter postoperative LOS did not have a higher probability of readmission than a patient with a longer postoperative LOS, independent of when during the study interval they were discharged. In other words, even though patients were older with greater severity of illness over the time period studied, the reduction of LOS was not associated with an increase in readmission.

14 Bohmer, Newell, and Torchiana

Surgery July 2002

Table III. Predictors (odds ratios and 95% confidence limits) of readmission Predictor

Odds ratio

LOS < 5 days LOS 5 to 7 days LOS 8 to 10 days LOS 11 to 15 days D/C home D/C home with services D/C to facility Male gender Age less than 56 years old Age 56 to 65 years old (ns) Age 66 to 75 years old Age 76 to 85 years old Age 86 to 95 years old (ns) Emergency procedure Urgent procedure (ns) Elective procedure CHF Renal failure Diabetes Atrial fibrillation

1.0 1.37 1.62 1.53 1.0 1.14 1.54 0.87 1.0 1.09 1.28 1.34 1.61 1.0 0.92 0.72 1.13 1.75 1.32 1.32

95% confidence limits — 1.16-1.61 1.34-1.96 1.23-1.90 — 1.01-1.29 1.33-1.78 0.78-0.96 — 0.93-1.26 1.11-1.48 1.13-1.59 0.97-2.67 — 0.81-1.05 0.64-0.80 1.01-1.28 1.45-2.12 1.20-1.46 1.19-1.47

Thirty-three other ICD-9 codes were entered as covariates and were found not to predict readmission. CHF, Congestive heart failure; D/C, discharged; ns, not significant.

DISCUSSION Our study was motivated by the conviction that excessively early discharge could harm patients. During the period of this study the postoperative LOS after bypass operation decreased by 19%, but 30-day readmission did not increase, in spite of increased age and acuity. This observation suggests that the decrease in postoperative LOS was not excessive and was driven by appropriate practice. On the other hand, readmission may not have changed because the threshold for admission increased during the same interval so that conditions that previously might have led to readmission were increasingly managed outside of the hospital. In support of this observation, the background admission rate of Massachusetts residents to Massachusetts acute hospitals for all causes was 2.7 per 100 residents at the beginning of this time period, falling monotonically and linearly to 2.4 per 100 at the end. Subjectively, it seems that both explanations are correct. There clearly has been both more aggressive early mobilization and discharge home, but patients seem to do better out of the hospital and when asked to do more. This philosophy has carried into outpatient management, because minor wound problems and atrial fibrillation with a well-controlled ventricular response are less likely to result in readmission. Finally, there has been a major increased use of alternate sites of care with less cost and lower intensity medical services.

This change has almost certainly helped keep the readmission rate down by managing basic social and medical issues that otherwise might be unmanageable at home. Decreased LOS has been accomplished in the name of efficiency, the intended incentive of prospective payment. Our data show that the potential savings from decreased LOS were not offset by increased readmissions. However, the use of home care services and postacute facilities increased substantially. This trend was more marked in the older Medicare population. In contrast to commercial insurers, which may vigorously manage referrals to postacute care, during the time of this study Medicare did not require previous approval for referral to visiting nurse services or rehabilitation facilities and paid these entities on a per diem rate that may have encouraged use and duration of these services. As a result, it is possible that whereas acute hospital costs have decreased, the costs to the health care delivery system as a whole have not also decreased because of the offsetting increase in home services and rehabilitation costs.8,9 Unfortunately, the data needed to make this calculation accurately are not available from the database we used. Our estimate is that the savings from the 1.4-day decrease in hospital LOS we observed is approximately equal to the additional costs experienced by discharging 12% more patients to post-

Bohmer, Newell, and Torchiana 15

Surgery Volume 132, Number 1

acute care facilities and increasing the use of home care services. The acute hospital saving is approximately $1400 per patient (if one approximates the cost of an acute hospital day at $1000 after postoperative day 4  1.4 fewer days of hospitalization). The comparable incremental cost of rehabilitation could be as much as $8600 (15.7 days average LOS at $550 per day) for 12% more patients, or $1036 per patient overall (to this would also be added the costs of additional use of home care services). This leaves the costs saved and the new costs incurred roughly equal, although there is arguably a great deal of imprecision in the numbers. The assumptions in this calculation are crude; although $1000 for a hospital day is a rough approximation of total cost, clearly the days at the end of a hospitalization are the least costly of all, and it takes operational changes to realize the full savings (closing beds, reducing staff). As a result, the cost savings may be overstated. Similarly, although the 15.7 days average LOS at the postacute facility and $550 payment are actual numbers from a local facility from the time period of this study, these are probably much less in today’s practice. Nonetheless, the exercise is valuable because it suggests once again that trying to cut costs in medicine is like squeezing a balloon; cost reductions in one place tend to reappear somewhere else. Our study has important limitations. There are inadequate clinical variables in administrative databases for effective risk adjustment, and it is possible that adequate severity measurements would lead to a different set of conclusions.10 Coding practices vary among institutions, and coding of clinical diagnoses may be incomplete.11 Readmission has limited value as an outcome measure,12 and there are patient inconveniences and hidden costs associated with early discharge that are not measured by readmission rate. Nevertheless, using the measure of 30-day readmission we found no evidence that shorter LOS at the levels observed disadvantage patients after CABG operation. At least in Massachusetts where CABG operations are limited to tertiary institutions, hospitals that monitor only their own readmission rates after heart operations as a measure of quality are missing half of the readmissions that go to other institutions. With decreasing LOS, we saw

a marked increase in the use of home services and postacute facilities. Patients having heart operations are getting older. Because the time for recovery after an operation cannot be compressed beyond a certain point, patients need to be cared for somewhere, and the judicious use of lower cost facilities is probably appropriate. We would like to thank the Commonwealth of Massachusetts Division of Health Care Finance and Policy for their assistance with the data, and Robert Seger, MBA, of the Massachusetts General Hospital Clinical Care Management Unit for his help with ICD-9 coding. REFERENCES 1. Reinhardt UE, Spending more through “cost control”: our obsessive quest to gut the hospital. Health Aff 1996;15: 145-54. 2. Engelman RM, Roussou JA, Flack JE III, Deaton DW, Humphrey CB, Ellison LH, et al. Fast-track recovery of the coronary bypass patient. Ann Thorac Surg 1994;58:1742-6. 3. Engelman RM. Mechanisms to reduce hospital stays. Ann Thorac Surg 1996;61:S26-9. 4. Krohn BG, Kay JH, Mendez MA, Zubiate P, Kay GL. Rapid sustained recovery after cardiac operations. J Thorac Cardiovasc Surg 1990;100:194-7. 5. Ferguson TB, Dziubian SW, Edwards FH, Eiken MC, Shroyer LW, Pairolero PC, et al. The STS national database: current changes and challenges in the new millennium. Ann Thorac Surg 2000;69:680-91. 6. Dixon JW, editor. BMDP Statistical Software Manual: To Accompany the 7.0 Software Release. Berkeley (CA): University of California Press; 1992. 7. Lahey, SJ, Campos CT, Jennings B, Pawlow P, Stokes T, Levitsky S. Hospital readmission after cardiac surgery. Does “fast track” cardiac surgery result in cost saving or cost shifting? Circulation 1998;98(Suppl):II35-40. 8. D’Agostino RS, Jacobson J, Clarkson M, Svensson LG, Williamson C, Shahian, DM. Readmission after cardiac operations: prevalence, patterns and predisposing factors. J Thorac Cardiovasc Surg 1999;118:823-32. 9. Cowper PA, Peterson ED, DeLong ER, Jollis JG, Muhlbaier LH, Mark DB. Impact of early discharge after coronary artery bypass graft surgery on rates of hospital readmission and death. J Am Coll Cardiol 1997;30:908-13. 10. Jollis JG, Romano PA. Pennsylvania’s focus on heart attack—grading the scorecard. N Engl J Med 1998;338: 983-7. 11. Jollis JG, Ancukiewicz M, DeLong ER, Pryor DB, Muhlbaier LH, Mark DB. Discordance of databases designed for claims payment versus clinic information systems. Ann Intern Med 1993;119:844-50. 12. Thomas JW. Does risk adjusted readmission rate provide valid information on hospital quality? Inquiry 1996;33:258-70.