- Email: [email protected]
Hospital Readmission by Method of Data Collection
Hospital Readmission by Method of Data Collection Elizabeth M Hechenbleikner, MD, Martin A Makary, MD, MPH, FACS, Daniel V Samarov, Jennifer L Bennett, BA, Susan L Gearhart, MD, FACS, Jonathan E Efron, MD, FACS, Elizabeth C Wick, MD, FACS
PhD,
Hospital readmissions are increasingly used to pay hospitals differently. We hypothesized that readmission rates, readmissions related to index admission, and potentially unnecessary readmissions vary by data collection method for surgical patients. STUDY DESIGN: Using 3 different data collection methods, we compared 30-day unplanned readmission rates and potentially unnecessary readmissions among colorectal surgery patients at a single institution between July 2009 and November 2011. We compared the NSQIP clinical reviewer method, the University HealthSystem Consortium (UHC) administrative billing data method, and physician medical record review. RESULTS: Seven hundred and thirty-five colorectal surgery patients were identified with readmission rates as follows: NSQIP 14.6% (107 of 735) vs UHC 17.6% (129 of 735). The NSQIP method identified 9 readmissions not found in billing records because the readmission occurred at another hospital (n ¼ 7) or due to a discrepancy in definition (n ¼ 2). The UHC method identified 31 readmissions not identified by NSQIP because of a broader readmission definition (n ¼ 20) or were missed by reviewers (n ¼ 11). The NSQIP method identified 72% of readmissions as related to index admission and physician chart review identified 83%. The UHC method identified 51% of readmissions as related to index admission and physician chart review identified 86%. Sixty-six of 129 UHC readmissions (51%) were deemed potentially preventable; based on physician chart review, 112 of 129 readmissions (87%) were deemed clinically necessary at the time of presentation. Most readmissions were due to surgical site infections (46 of 129 [36%]) and dehydration (30 of 129 [23%]). With improved patient-care efforts, 41 of 129 (31.8%) complications might not have required readmission. CONCLUSIONS: Readmission rates and unnecessary readmissions vary depending on data collection methodology. Reimbursements based on readmission should use standardized and fair methods to minimize perverse incentives that penalize hospitals for appropriate care of high-risk surgical patients. (J Am Coll Surg 2013;216:1150e1158. 2013 by the American College of Surgeons)
BACKGROUND:
The Medicare Payment Advisory Commission reported that 17.6% of index hospital admissions are associated with a readmission within 30 days of discharge. The Medicare Payment Advisory Commission has several definitions for potentially preventable readmissions, including those Disclosure Information: Nothing to disclose. Presented at the Annual Meeting of the Maryland Chapter of the American College of Surgeons and the Annual Meeting, Resident Research Forum, Baltimore, MD, November 2012. Received November 23, 2012; Revised January 10, 2013; Accepted January 25, 2013. From the Department of Surgery, Johns Hopkins University, Baltimore (Hechenbleikner, Makary, Bennett, Gearhart, Efron, Wick) and National Institute of Standards and Technology, Gaithersburg (Samarov), MD. Correspondence address: Elizabeth C Wick, MD, FACS, Department of Surgery, Johns Hopkins University School of Medicine, Blalock Room 658, 600 N Wolfe St, Baltimore, MD 21287. email: [email protected]
ª 2013 by the American College of Surgeons Published by Elsevier Inc.
that could have been avoided with improved index hospitalization patient care, discharge planning, or outpatient care coordination.1 Currently, there is no consensus on the best methodology for establishing potentially preventable readmissions and, by default, pay-for-performance incentives are beginning to use all-cause readmission rates. Starting in October 2012, two thirds of US hospitals were penalized for high all-cause readmission rates among patients with index admissions for acute myocardial infarction, heart failure, and pneumonia.2 In total, it is predicted that hospitals will forfeit about $280 million in Medicare funds. This is likely to be extended to a hospital wide allcause unplanned readmission measure starting in 2013. Although measuring quality is an important goal, surgical patients can be different from medical patients. In a large study of Medicare beneficiaries, most 30-day
1150
ISSN 1072-7515/13/$36.00 http://dx.doi.org/10.1016/j.jamcollsurg.2013.01.057
Hechenbleikner et al
Vol. 216, No. 6, June 2013
Abbreviations and Acronyms
CMS OR SSI UHC
¼ ¼ ¼ ¼
Centers for Medicare and Medicaid Services odds ratio surgical site infection University HealthSystem Consortium
readmissions after surgery were related more to a patient’s underlying medical condition than to the operation itself, with cardiac stent placement and gastrointestinal disorders being the leading diagnoses at readmission.3 Although some surgical complications might be preventable, some are inherent risks associated with the procedure and might be intrinsically associated with a high readmission rate as a part of safe management. Therefore, defining preventable or unnecessary surgical readmissions is a challenge to ensure fair measurements of quality. We hypothesize that administrative data might not appropriately distinguish preventable readmissions from nonpreventable readmissions at a hospital level. To address this question, we designed a study comparing a clinical registry, an administrative database, and a clinical case review by a surgeon with the following major aims: to analyze the variation in readmission rates and readmission diagnoses by data collection method, to identify the subset of clinically unnecessary readmissions, and to determine which complications leading to readmission might have been prevented with improved patient-care efforts.
METHODS Patient population All patients who underwent colon or rectal resections between July 2009 and November 2011 were identified by Current Procedural Terminology codes: 44140-147, 44150-151, 44155-158, 44160, 44204-208, 44210-212, 44130, 44395, 44397, 44402, 44113, and 44550. The American College of Surgeons’ NSQIP data were supplemented with additional data abstracted from patient charts. Table 1.
Readmissions and Data Collection
1151
National Surgical Quality Improvement Program Johns Hopkins Hospital participates in the targeted procedure module of NSQIP with 100% capture of all colon and rectal cases. This program was initiated in 2009 and 30-day readmissions are tracked. All NSQIP data at our hospital are collected by 1 full-time nurse reviewer and 2 part-time nurse reviewers. All reviewers are experienced and have been audited by the NSQIP program in the past and found to be valid. Data are abstracted from hospital electronic medical records and patient follow-up phone calls. The NSQIP defines the 30-day readmission window as a readmission occurring within 30 days of the index surgical procedure date. Readmissions to our hospital or any other hospital are included. Planned readmissions are also included in NSQIP’s registry. Principal readmission diagnosis and relationship between index operation and readmission are determined by the nurse reviewers (Table 1). Reviewers do not indicate if the readmission was considered potentially preventable at the time of presentation. University HealthSystem Consortium University HealthSystem Consortium (UHC) is an alliance of 116 academic medical centers and 275 affiliated hospitals that reports risk-adjusted performance metrics to its member institutions based on administrative data. The UHC defines the 30-day readmission window as a readmission occurring within 30 days of the index admission discharge date, similar to the proposed Medicare measure. The readmission is assigned a primary diagnosis code using the following algorithms: diagnosis-related group (Centers for Medicare and Medicaid Services [CMS]), all patient refined diagnosis-related group (3M), Clinical Classification System Category (Agency for Healthcare Research and Quality), and ICD-9 complication codes. If the readmission diagnosis code is any of the following it is considered planned and ultimately excluded: scheduled chemotherapy, radiation therapy, or dialysis treatment; same-day transfer to psychiatric facility, oncology ward, or inpatient rehabilitation; alcohol and
30-Day Readmission Criteria by Method
NSQIP criteria
Nurse reviewer medical record interpretation and/or patient interview
University HealthSystem Consortium Criteria
If any of the following criteria are met: Index hospitalization DRG ¼ readmission DRG Index hospitalization APR-DRG ¼ readmission APR-DRG Principal readmission diagnosis is a complication code (ICD-9 codes 996.00e999.9) Clinical classification system category of index hospitalization principal diagnosis ¼ clinical classification system category of readmission principal diagnosis Clinical classification system category of index hospitalization primary procedure ¼ clinical classification system category of readmission primary procedure
APR-DRG, all patient refined diagnosis-related group; DRG, diagnosis-related group.
1152
Hechenbleikner et al
Readmissions and Data Collection
drug treatment admission; and labor and delivery admission. Planned readmissions for any other reason are not excluded. Based on these 4 diagnosis code algorithms, a readmission is identified as related or unrelated to the index admission; all related readmissions are considered potentially preventable. A readmission is defined as related when certain index admission diagnosis codes match readmission diagnosis codes (Table 1). Surgeon case review All readmissions identified based on the 2 methodologies (NSQIP and UHC) were confirmed by physician chart review. The relationship to index operation or admission was determined clinically based on the patient’s presenting symptoms, laboratory data, and imaging studies. Readmissions were then evaluated for whether they were necessary at the time of the postoperative patient’s presentation. Readmissions were considered clinically necessary at the time of presentation if patients met one or more of the following criteria: ICU admission; unplanned return to the operating room; nonoperating room invasive procedure; placement of peripherally inserted central catheter; and abnormal vital signs, laboratory results, or radiographic data as defined in Table 2. Readmissions that did not meet any of these criteria were considered potentially unnecessary from a clinical standpoint. The primary readmission diagnosis or underlying complication was then assessed to determine whether it occurred during the index admission or after hospital discharge. Finally, if there was an underlying complication that led to the 30-day readmission, it was assessed for preventability based on clinical judgment to determine whether it could have been avoided during the index hospital stay and/or with enhanced outpatient care efforts. Statistical analysis A multivariate logistic regression analysis for risk factors predictive of readmission was performed using the R statistical programming language (R Foundation for Statistical Computing). Risk factors identified as contributing substantially to readmission were selected using a stepwise procedure.4 Patient-specific, intraoperative, and postoperative variables (Table 3) from the NSQIP database were included in the regression analysis. This study was approved by the Institutional Review Board at Johns Hopkins Hospital.
RESULTS We identified 735 patients who underwent colorectal operations between July 2009 and November 2011. Median age was 56 years (range 19 to 96 years). Six hundred and
J Am Coll Surg
Table 2. Criteria for Clinically Necessary 30-Day Readmission at the Time of the Patient’s Presentation At the time of rehospitalization, one or more of the following clinical criteria must be met: Major criteria ICU admission Unplanned return to the operating room Nonoperating room invasive procedure* Peripherally inserted central catheter (ie, for hydration, antibiotics, etc) Minor criteria Abnormal vital signs Temperature >38.3 C or <36 C Tachycardia (heart rate 110 bpm) Absolute hypotension (systolic blood pressure <90 mmHg) Clinical documentation of orthostatic hypotension Acute renal failure defined as 0.5 mg/dL increase in serum creatinine from a baseline of 1.9 mg/dL 1.0mg/dL increase in serum creatinine from a baseline of 2.0e4.9 mg/dL 1.5 mg/dL increase in serum creatinine from a baseline of 5.0 mg/dL White blood cell count >12,000 or <4,000 cells/mm3 Severe electrolyte imbalances or hypoglycemia requiring treatment Drop in hemoglobin count requiring blood transfusion CT scan evidence of bowel obstruction or anastomotic leak Bowel obstruction requiring nasogastric tube placement or parenteral nutrition *Percutaneous abdominal drain, endoscopy, bronchoscopy, cardiac catheterization, biliary drain placement, etc.
thirty-four (86%) patients underwent colon resections and 101 (14%) underwent rectal resections. Two hundred and ninety-four patients had ostomies created as part of their index operation (219 ileostomies, 75 colostomies). Two hundred and forty-one procedures (33%) were laparoscopic and 73 (10%) operations were done emergently. The most common indications for the procedures were 228 colorectal cancer (31%), 181 inflammatory bowel disease (24.6%), and 45 diverticulitis (6.1%; Table 3). The most common morbidity in the NSQIP database was surgical site infection (SSI) with 166 patients (22.5%) being diagnosed within 30 days of surgery; most SSIs were superficial incisional (116 patients [15.8%]) followed by organ space (44 patients [6%]). Significant risk factors for readmission in this group of patients were body mass index (odds ratio [OR] ¼ 1.05; p ¼ 0.001), presence of an ostomy (OR ¼ 2.59; p < 0.001), superficial incisional SSI (OR ¼ 1.74; p ¼ 0.043), deep incisional SSI (OR ¼ 14.78; p ¼ 0.004), and organ space SSI (11.54; p < 0.001). The 30-day unplanned readmission rate in NSQIP was 107 of 735 patients (14.6%) and in UHC was 129 of 735 patients (17.6%; Fig. 1). Median time to readmission was
Hechenbleikner et al
Vol. 216, No. 6, June 2013
Table 3.
Readmissions and Data Collection
1153
Characteristics of Readmitted and Nonreadmitted Patients
Characteristics
Patient-specific* Age, y, mean (range) Male sex, n (%) BMI, mean (range) Diabetes mellitus, n (%) Current smoker, n (%) Steroid use, n (%) Albumin, g/dL, mean (range) Emergent case, n (%) Surgical indication, n (%) Colorectal cancer (n ¼ 228) Diverticulitis (n ¼ 45) Inflammatory bowel disease (n ¼ 181) Other cancer (n ¼ 32) Miscellaneous (n ¼ 249)y Intraoperative, n (%) Type of resection Colon (n ¼ 634) Rectal (n ¼ 101) Type of approach Open (n ¼ 494) Laparoscopic (n ¼ 241) Presence of ostomy, n (%) Yes No Length of operation, h, mean (range) Postoperative, n (%) SSI Superficial incisional SSI Deep incisional SSI Organ space SSI Length of stay, d, median (range) 30-d mortality
Readmitted patients (n ¼ 129)
57 67 29 22 24 23 3.8 15 39/228 16/45 31/181 9/32 34/249
(22e86) (52) (15e59) (17) (18.6) (17.8) (1.7e5.2) (11.6) (17) (35.6) (17) (28) (14)
Nonreadmitted patients (n ¼ 606)
56 285 27 50 98 91 3.9 58 189/228 29/45 150/181 23/32 215/249
(19e96) (47) (13e58) (8.3) (16.2) (15) (1.1e5.1) (9.6) (83) (64.4) (83) (72) (86)
p Value
Odds ratio
NS NS 0.001 NS NS NS NS NS
NS NS 1.05 NS NS NS NS NS
0.002 NS 0.001 NS <0.001
0.35 NS 0.32 NS 0.24
108/634 (17) 21/101 (21)
526/634 (83) 80/101 (79)
NS NS
NS NS
96/494 (19) 33/241 (14)
398/494 (81) 208/241 (86)
NS NS
NS NS
74 (57) 55 (43) 4.9 (1.5e12.2)
220 (36) 386 (64) 4.2 (1e14)
<0.001 NS NS
2.59 NS NS
58 27 4 27 9 1
108 89 2 17 7 19
<0.001 0.043 0.004 <0.001 NS NS
3.39 1.74 14.78 11.54 NS NS
(45) (21) (3.1) (21) (3e65) (0.8)
(18) (14.7) (0.3) (2.8) (1e153) (3.1)
*All data for patient-specific and intraoperative variables were collected at the time of the index operation. y Miscellaneous includes multiple indications, such as volvulus, rectal prolapse, benign neoplasm, vascular insufficiency, accidental perforation, diverticulosis, etc. BMI, body mass index (calculated as kg/m2); SSI, surgical site infection.
8 days (range 1 to 30 days) and median length of hospitalization after readmission was 5 days (range 1 to 53 days). The NSQIP registry included 1 patient with a planned readmission for a right hemicolectomy. The UHC readmission cohort included 3 patients with planned readmissions (portal vein embolization, ostomy reversal, and right hemicolectomy). All planned readmissions were excluded from our analysis but were ultimately reported by NSQIP and UHC, respectively. Although the majority of unplanned readmissions (98 patients) were identified in both methodologies, multiple discrepancies were noted. The NSQIP method identified 9 readmissions not found with the UHC method because the
readmission occurred at another hospital (7 patients) or met UHC exclusion criteria (1 patient was a psychiatric facility transfer; 1 patient was an inpatient rehabilitation transfer). The UHC method identified 31 readmissions not found in NSQIP; 20 were because of the discrepancy in the 30-day readmission window definition (UHC defines the 30-day window as beginning at the date of discharge as opposed to the date of index operation) and 11 because they were missed by clinical reviewers. Next, we examined the diagnosis associated with readmission and whether the readmission was related to the index operation or admission based on NSQIP (nurse reviewer determination) and UHC criteria (diagnosis
1154
Hechenbleikner et al
Readmissions and Data Collection
J Am Coll Surg
Figure 1. Thirty-day readmission rates as determined by NSQIP and University HealthSystem Consortium (UHC) methods.
code algorithm), respectively. The NSQIP reviewers evaluate the relationship between a readmission and the index operation but do not currently indicate whether the readmission was considered preventable. Based on UHC methodology, if the readmission is related to the index admission, the readmission is deemed potentially preventable. These determinations were compared with surgeon chart review to assess accuracy. The principal diagnoses or complications that led to the readmission were divided into the following categories: bleeding, dehydration, infection, gastrointestinal related, and other (Table 4). In the NSQIP registry, the percentage of readmissions related to index surgical procedure was 72%. In the UHC registry, the percentage of readmissions related to index hospitalization was 51%. Surgeon case review demonstrated higher rates of readmissions related to index operation or admission at 83% and 86%, respectively (Table 4); the majority of readmissions were due to postoperative complications. The most frequent complications requiring readmission were SSI (46 of 129 patients [36%]), dehydration with high ileostomy output (20 of 129 patients [15.5%]), and gastrointestinal issues, including partial bowel obstruction or paralytic ileus (17 of 129 patients [13%]). In general, UHC consistently reported lower rates of readmissions related to index hospital stay compared with other methods. For example, UHC reported that 10% of readmissions for dehydration were related to the index admission as compared with surgeon case review at 97% (Table 4).
Identification of readmissions for infection as related to index operation or admission was most consistent in the 2 methodologies, 93% NSQIP vs 84% UHC, respectively. In addition, criteria were established to determine if, at the time the patient presented for readmission, this admission was clinically necessary for safe patient care (Table 2). By surgeon review, we established that 17 of 129 (13%) readmissions were potentially preventable from a clinical perspective; none of these patients met either major or minor criteria summarized in Table 2. The clinically avoidable readmissions were in the following categories: superficial SSI (5 patients), mild dehydration (4 patients), gastrointestinal-related issues (5 patients), and 3 other patients (skin excoriation, kidney stones, and failure to thrive). One hundred and 12 readmissions (87%) were determined to be clinically necessary at the time of presentation; 61 patients (54%) met one or more major criteria, 51 patients (46%) met one or more minor criteria alone (Table 5). At the time of presentation, 26 of 30 (87%) readmissions for complications of dehydration were not avoidable. The majority (22 of 26 [85%]) of these patients, had an ostomy (20 ileostomy, 2 colostomy). Of the 46 readmissions for SSI, 41 (89%) were necessary for safe patient care; many of these patients had abnormal vital signs or significant leukocytosis at time of presentation or required invasive procedures, ICU admissions, and/or surgical exploration during readmission. Eleven patients required ICU admission for severe dehydration (n ¼ 5) or organ
Vol. 216, No. 6, June 2013
Hechenbleikner et al
Readmissions and Data Collection
1155
Table 4. Accuracy of 30-Day Readmissions Related to Index Operation or Admission in NSQIP and University HealthCare System
Patient groups
NSQIP All patients Principal readmission diagnosis categoryy Bleeding Dehydration Infection Gastrointestinal Other University HealthCare System All patients Principal readmission diagnosis category Bleeding Dehydration Infection Gastrointestinal Other
Total no. of readmissions
Readmissions related to index operation n %
Surgeon case review n %
71*
51
72
59
83
3 16 27 16 9
0 9 25 11 6
0 56 93 69 67
2 15 25 13 4
67 94 93 81 44
129
66
51
111
86
4 30 51 27 17
1 3 43 14 5
25 10 84 52 29
3 29 46 25 8
75 97 90 93 47
*NSQIP began tracking readmission relatedness to index operation in spring 2010; only 71 patients were reviewed. y Readmission categories: bleeding (ie, gastrointestinal hemorrhage); other (ie, failure to thrive, medication overdose); gastrointestinal (ie, ileus, bowel obstruction); dehydration (ie, high ostomy output); and infection (urinary tract infection, surgical site infection).
space SSI (n ¼ 6) and 8 patients required reoperation for organ space SSI (n ¼ 5) or deep incisional SSI (n ¼ 3). Overall, 39 patients required percutaneous interventional and/or endoscopic procedures; 20 SSI patients (19 organ space, 1 deep incisional) required percutaneous abdominal drainage. Among all 129 patients who were readmitted, in 15 patients (11.6%) a complication was diagnosed during the index admission compared with 114 patients (88.4%) who were diagnosed with a complication at the time of readmission (data not shown). Complications diagnosed during the index admission were dehydration in 4 patients; gastrointestinal-related in 3 patients; SSIs in 7 patients; and deep venous thrombosis in 1 patient. Complications diagnosed at readmission were bleeding in 4 patients; dehydration in 26 patients; gastrointestinal-related in 24 patients; pneumonia in 1 patient; diverticulitis in 1 patient; colitis in 1 patient; cellulitis in 1 patient; SSI in 39 patients; urinary tract infection in 1 patient; and other (ie, myocardial infarction, kidney stones) in 16 patients. Among the 15 patients diagnosed with complications at the time of index admission, 10 of the patients (67%) had complications that could have been potentially avoided with different treatment regimens. Readmissions for dehydration with high ostomy output (n ¼ 4) might have been avoided with enhanced patient education, medication reconciliation, and/or outpatient care coordination. Organ space SSIs with anastomotic leak (n ¼ 3) requiring readmission might have been prevented if an
ostomy had been created during the index operation. One organ space SSI (abscess, no anastomotic leak) readmission might have been prevented with better inpatient care; this patient had a persistent intra-abdominal fluid collection after percutaneous abdominal drainage, which could have been drained again before discharge. Two patients with superficial SSIs were readmitted 1 day after hospital discharge. In this case, both improved inpatient efforts, including aggressive local wound care, and enhanced care transitions might have prevented these readmissions. Among the 114 complications diagnosed at readmission, 31 (27%) might not have required rehospitalization with enhanced patient-care efforts. Twentytwo dehydration readmissions (16 high ostomy output, 3 diarrhea, 1 vomiting, 1 poor fluid intake, and 1 misuse of diuretics), 5 gastrointestinal-related readmissions (1 severe constipation, 1 leakage around ostomy, 2 diarrhea, and 1 colostomy retraction), and 4 other readmissions (2 skin excoriation, 1 hyponatremia, and 1 misuse of narcotics) might have been prevented with improved outpatient care coordination. Overall, readmissions for SSIs and other infectious complications diagnosed after hospital discharge were very difficult to assess for preventability because of the sparse documentation in the medical record between index admission and rehospitalization.
DISCUSSION Hospital readmissions are emerging as a leading quality indicator and prevention has been targeted as a strategy
1156
Table 5.
Hechenbleikner et al
J Am Coll Surg
Readmissions and Data Collection
Surgeon Case Review of 30-Day Readmissions
Clinically necessary vs unnecessary 30-d readmissions
Bleeding (n ¼ 4 [3%]) n %
Necessary readmissions Major criteria ICU admission Unplanned return to operating room Nonoperating room invasive procedure Peripherally inserted central catheter Minor criteria (alone) Potentially unnecessary readmissions
4 3 2 0 3 0 1 0
100 75 50 75 25
Dehydration (n ¼ 30 [23%]) n %
26 11 5 1 3 5 15 4
to improve patient safety and reduce health care expenditures. Beginning in October 2012, CMS started paying hospitals 1% less for excess readmissions for patients with 3 common diagnoses (ie, acute myocardial infarction, heart failure, pneumonia). The same provision will increase the penalty to 2% in September 2013 and 3% in September 2014. This rule is expected to be extended to a hospital-wide all-cause unplanned readmission measure in 2013 that will include surgical patients. The CMS comprehensive all-cause, hospital-wide readmission measure will identify patients with unplanned readmissions within 30 days of discharge.5 To set benchmarks for unplanned, related surgical readmissions, the hospital-wide readmission measure will use algorithms matching procedure codes and discharge diagnosis categories from administrative claims data, similar to UHC methodology. Our study of readmission definitions and principal readmission diagnoses using administrative data and NSQIP as compared with physician chart review revealed multiple discrepancies. Although administrative data identified an unplanned readmission rate of 17.6%, NSQIP found the rate to be 14.6%. More importantly, we found that administrative data were not able to consistently determine whether a readmission was related to the index hospitalization. Although physician case review determined most readmissions were related to postoperative complications, administrative data frequently failed to relate the unplanned readmissions to the index admission or identify a postoperative complication as the reason for readmission. The accuracy varied based on the postoperative complication. Administrative data were most accurate at relating postoperative infections like SSI and least accurate at relating dehydration to the index hospitalization. We postulate that this is because postoperative infection is frequently coded with an ICD-9 complication code and this is automatically identified as related to index admission; in contrast, dehydration, the second
87 42.3 19.2 3.8 11.5 19.2 57.7 13
Infection (n ¼ 51 [40%]) n %
46 32 6 8 21 6 14 5
90 70 13 17.4 45.7 13 30 10
Gastrointestinal (n ¼ 27 [21%]) n %
22 8 1 2 5 3 14 5
81.5 36.4 4.5 9.1 22.7 13.6 63.6 18.5
Other (n ¼ 17 [13%]) n %
14 7 2 1 5 1 7 3
82 50 14.3 7.1 35.7 7.1 50 18
Totals (n ¼ 129) n %
112 61 16 12 37 15 51 17
87 54 14.2 10.7 33 13.4 46 13
leading cause of readmissions in this population, is not. Depending on the type of surgical patient (colorectal vs vascular vs spinal), administrative data could markedly vary in their ability to establish the relationship between index admission and readmission but current models do not factor in case-mix adjustment. Additional refinement of these algorithms is needed before administrative data can be considered for establishing benchmarks that will impact financial reimbursement policies. Emerging evidence suggests that surgical complications are an important factor in postoperative hospital readmissions.6 Studies of general surgery patients have suggested that readmissions in this population are often the result of operative complications, including SSI and bowel obstruction, as opposed to an exacerbation of underlying medical conditions.7-9 In surgical patients with postoperative complications, readmission is frequently clinically necessary for safe care. In our study, administrative data (UHC) predominantly defined readmissions as potentially preventable by identifying the principal readmission diagnosis as a complication code. The majority of preventable readmissions identified by administrative data in our study were related to postoperative infections. Based on clinical judgment, we determined that most of these readmissions were necessary for safe patient care at the time of presentation; 45.7% and 17.4% of patients required invasive procedures or reoperation, respectively (Table 5). Additional study is needed to determine if SSIs are preventable in colorectal surgery. We did, however, identify several postoperative complications (organ space and superficial SSIs) diagnosed during index admission that might have not required rehospitalization with enhanced inpatient treatment. Unfortunately, all-cause readmission policies like those adopted by CMS might have the perverse effect of incentivizing physicians not to accept patients at high risk for readmission or not to readmit postoperative patients when medically indicated. In other areas, implementation of hospital-based pay-for-performance programs and
Vol. 216, No. 6, June 2013
Hechenbleikner et al
benchmarks has demonstrated conflicting evidence to date for improving patient outcomes such as mortality.10,11 In colorectal surgery, prevention of readmission for an acutely ill postoperative patient could delay appropriate care and potentially be harmful. Accordingly, resources allocated toward preventing surgical readmissions should be redirected toward individual and systems-based practices for preventing complications. We previously12 reported on the implementation of an interdisciplinary program (CUSP [Comprehensive Unit-Based Safety Program]) to decrease SSI rates in a high-risk surgical cohort by standardizing preoperative and intraoperative techniques. Hospital investment in inpatient quality-improvement programs might be the most effective strategy for preventing surgical readmissions. We did identify, by physician chart review, a subset of postoperative complications that might not have required readmission with enhanced outpatient care coordination. The majority of these readmissions were for patients with dehydration and high ostomy output but also included constipation, skin excoriation, and misuse of narcotics. Although no care-coordination interventions have focused on surgical patients, there are reports of effective models in medical patients. Most studies have focused on transitional care, such as discharge planning, medication reconciliation, and timely outpatient follow-up. In a 2011 systematic review of 43 studies, the most common interventions tested were predischarge patient education, discharge planning, and/or postdischarge telephone calls.13 Only 5 randomized trials showed a significant absolute risk reduction in 30-day readmissionsd1 with a single intervention and 4 with multiple or “bundled” interventions. The authors concluded that no single intervention or bundled interventions appeared to consistently decrease 30-day readmissions and that additional research is required. Centers have begun similar interventions in surgical patients with aggressive follow-up in the postdischarge postoperative period. These efforts include frequent calls to the patient and early office visits with a nurse or physician. More study is needed to evaluate the efficacy of such interventions in this high-risk population. There were several important limitations to this study. First, this was a retrospective cohort study with a small number of patients. Second, varying definitions used by NSQIP and UHC for determining 30-day readmissions and the relationship to index admission made qualitative comparisons difficult. Third, there are no established guidelines in the literature for determining preventable or clinically unnecessary hospital readmissions in surgical patients. Finally, our analysis is based on a select, high-risk surgical cohort at a single tertiary referral academic center and might not be representative of readmissions at other institutions.
Readmissions and Data Collection
1157
CONCLUSIONS Financial penalties placed on hospitals for excess readmissions are imminent and are currently based on administrative data. All-cause readmission benchmarks that are derived from administrative data should be used cautiously in surgical patients until additional validation and research has been performed. Within a select surgical cohort, we demonstrate that the ability to relate index and readmitting conditions and clinically unnecessary readmissions is more reliable when using surgeon-endorsed definitions and an independent clinical reviewer. In our sample, SSI was the only postoperative complication that was consistently linked back to the index admission and the majority of these readmissions were necessary for safe patient care, based on physician chart review. Frequently, these patients required invasive procedures or reoperation, so any delay of treatment could have resulted in considerable harm. With the addition of well-defined, consistent criteria, prospective clinical registries such as NSQIP that use independent clinical reviewers might be able to track postoperative complications driving readmissions and ultimately determine which interventions can help reduce readmissions. If most surgical readmissions are due to complications that require treatment and hospitalization at the time of presentation, the best way to decrease readmissions is to prevent complications from ever happening. Future research and resources should focus on perioperative interventions to decrease surgical complications and how these interventions compare with other efforts, such as outpatient care coordination, to decrease 30-day readmissions. Author Contributions Study conception and design: Hechenbleikner, Makary, Gearhart, Efron, Wick Acquisition of data: Hechenbleikner, Bennett, Wick Analysis and interpretation of data: Hechenbleikner, Samarov, Wick Drafting of manuscript: Hechenbleikner, Makary, Samarov, Bennett, Gearhart, Efron, Wick Critical revision: Hechenbleikner, Makary, Samarov, Bennett, Gearhart, Efron, Wick Acknowledgment: The authors thank Lucy Mitchell, RN, MA, QI Specialist, and SCNR ACS NSQIP, for assistance with acquisition, analysis, and interpretation of data as well as Jason Miller, Director of Clinical Analytics, for assistance with UHC data. REFERENCES 1. Medicare Payment Advisory Commission. Report to Congress: promoting efficiency in Medicare, Chapter 5. Available at: www.medpac.gov. Accessed June 15, 2007.
1158
Hechenbleikner et al
Readmissions and Data Collection
2. Rau J. Medicare to penalize 2,211 hospitals for excess readmissions. Kaiser Health News. Available at: www.kaiserhealthnews. org. Accessed August 13, 2012. 3. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med 2009;360:1418e1428. 4. Hastie TJ, Pregibon D. Generalized linear models. In: Chambers JM, Hastie TJ, eds. Statistical Models in S. Pacific Grove, CA: Wadsworth & Brooks/Cole; 1991:195e248. 5. Horwitz L, Partovian C, Lin Z, et al. Hospital-wide (all-condition) 30-day risk-standardized readmission measure, draft measure methodology report. Yale New Haven Health Services Corporation/Center for Outcomes Research and Evaluation (YNHHSC/CORE). Available at: www.cms.gov. Accessed Aug 10, 2011. 6. Kassin MT, Owen RM, Perez SD, et al. Risk factors for 30-day hospital readmission among general surgery patients. J Am Coll Surg 2012;215:322e330. 7. Reddy DM, Townsend CM Jr, Kuo YF, et al. Readmission after pancreatectomy for pancreatic cancer in Medicare
8.
9. 10. 11. 12. 13.
J Am Coll Surg
patients. J Gastrointest Surg 2009;13:1963e1974; discussion 1974e1975. Schneider EB, Hyder O, Brooke BS, et al. Patient readmission and mortality after colorectal surgery for colon cancer: impact of length of stay relative to other clinical factors. J Am Coll Surg 2012;214:390e398; discussion 398e399. Ozturk E, Kiran RP, Remzi F, Fazio VW. Early readmission after ileoanal pouch surgery. Dis Colon Rectum 2009;52:1848e1853. Jha AK, Joynt KE, Oray EJ, et al. The long-term effect of premier pay for performance on patient outcomes. N Engl J Med 2012;366:1606e1615. Sutton M, Nikolova S, Boaden R, et al. Reduced mortality with hospital pay for performance in England. N Engl J Med 2012;367:1821e1828. Wick EC, Hobson DB, Bennett JL, et al. Implementation of a surgical comprehensive unit-based safety program to reduce surgical site infections. J Am Coll Surg 2012;215:193e200. Hansen LO, Young RS, Hinami K, et al. Interventions to reduce 30-day rehospitalization: a systematic review. Ann Intern Med 2011;155:520e528.
PhD,
Hospital readmissions are increasingly used to pay hospitals differently. We hypothesized that readmission rates, readmissions related to index admission, and potentially unnecessary readmissions vary by data collection method for surgical patients. STUDY DESIGN: Using 3 different data collection methods, we compared 30-day unplanned readmission rates and potentially unnecessary readmissions among colorectal surgery patients at a single institution between July 2009 and November 2011. We compared the NSQIP clinical reviewer method, the University HealthSystem Consortium (UHC) administrative billing data method, and physician medical record review. RESULTS: Seven hundred and thirty-five colorectal surgery patients were identified with readmission rates as follows: NSQIP 14.6% (107 of 735) vs UHC 17.6% (129 of 735). The NSQIP method identified 9 readmissions not found in billing records because the readmission occurred at another hospital (n ¼ 7) or due to a discrepancy in definition (n ¼ 2). The UHC method identified 31 readmissions not identified by NSQIP because of a broader readmission definition (n ¼ 20) or were missed by reviewers (n ¼ 11). The NSQIP method identified 72% of readmissions as related to index admission and physician chart review identified 83%. The UHC method identified 51% of readmissions as related to index admission and physician chart review identified 86%. Sixty-six of 129 UHC readmissions (51%) were deemed potentially preventable; based on physician chart review, 112 of 129 readmissions (87%) were deemed clinically necessary at the time of presentation. Most readmissions were due to surgical site infections (46 of 129 [36%]) and dehydration (30 of 129 [23%]). With improved patient-care efforts, 41 of 129 (31.8%) complications might not have required readmission. CONCLUSIONS: Readmission rates and unnecessary readmissions vary depending on data collection methodology. Reimbursements based on readmission should use standardized and fair methods to minimize perverse incentives that penalize hospitals for appropriate care of high-risk surgical patients. (J Am Coll Surg 2013;216:1150e1158. 2013 by the American College of Surgeons)
BACKGROUND:
The Medicare Payment Advisory Commission reported that 17.6% of index hospital admissions are associated with a readmission within 30 days of discharge. The Medicare Payment Advisory Commission has several definitions for potentially preventable readmissions, including those Disclosure Information: Nothing to disclose. Presented at the Annual Meeting of the Maryland Chapter of the American College of Surgeons and the Annual Meeting, Resident Research Forum, Baltimore, MD, November 2012. Received November 23, 2012; Revised January 10, 2013; Accepted January 25, 2013. From the Department of Surgery, Johns Hopkins University, Baltimore (Hechenbleikner, Makary, Bennett, Gearhart, Efron, Wick) and National Institute of Standards and Technology, Gaithersburg (Samarov), MD. Correspondence address: Elizabeth C Wick, MD, FACS, Department of Surgery, Johns Hopkins University School of Medicine, Blalock Room 658, 600 N Wolfe St, Baltimore, MD 21287. email: [email protected]
ª 2013 by the American College of Surgeons Published by Elsevier Inc.
that could have been avoided with improved index hospitalization patient care, discharge planning, or outpatient care coordination.1 Currently, there is no consensus on the best methodology for establishing potentially preventable readmissions and, by default, pay-for-performance incentives are beginning to use all-cause readmission rates. Starting in October 2012, two thirds of US hospitals were penalized for high all-cause readmission rates among patients with index admissions for acute myocardial infarction, heart failure, and pneumonia.2 In total, it is predicted that hospitals will forfeit about $280 million in Medicare funds. This is likely to be extended to a hospital wide allcause unplanned readmission measure starting in 2013. Although measuring quality is an important goal, surgical patients can be different from medical patients. In a large study of Medicare beneficiaries, most 30-day
1150
ISSN 1072-7515/13/$36.00 http://dx.doi.org/10.1016/j.jamcollsurg.2013.01.057
Hechenbleikner et al
Vol. 216, No. 6, June 2013
Abbreviations and Acronyms
CMS OR SSI UHC
¼ ¼ ¼ ¼
Centers for Medicare and Medicaid Services odds ratio surgical site infection University HealthSystem Consortium
readmissions after surgery were related more to a patient’s underlying medical condition than to the operation itself, with cardiac stent placement and gastrointestinal disorders being the leading diagnoses at readmission.3 Although some surgical complications might be preventable, some are inherent risks associated with the procedure and might be intrinsically associated with a high readmission rate as a part of safe management. Therefore, defining preventable or unnecessary surgical readmissions is a challenge to ensure fair measurements of quality. We hypothesize that administrative data might not appropriately distinguish preventable readmissions from nonpreventable readmissions at a hospital level. To address this question, we designed a study comparing a clinical registry, an administrative database, and a clinical case review by a surgeon with the following major aims: to analyze the variation in readmission rates and readmission diagnoses by data collection method, to identify the subset of clinically unnecessary readmissions, and to determine which complications leading to readmission might have been prevented with improved patient-care efforts.
METHODS Patient population All patients who underwent colon or rectal resections between July 2009 and November 2011 were identified by Current Procedural Terminology codes: 44140-147, 44150-151, 44155-158, 44160, 44204-208, 44210-212, 44130, 44395, 44397, 44402, 44113, and 44550. The American College of Surgeons’ NSQIP data were supplemented with additional data abstracted from patient charts. Table 1.
Readmissions and Data Collection
1151
National Surgical Quality Improvement Program Johns Hopkins Hospital participates in the targeted procedure module of NSQIP with 100% capture of all colon and rectal cases. This program was initiated in 2009 and 30-day readmissions are tracked. All NSQIP data at our hospital are collected by 1 full-time nurse reviewer and 2 part-time nurse reviewers. All reviewers are experienced and have been audited by the NSQIP program in the past and found to be valid. Data are abstracted from hospital electronic medical records and patient follow-up phone calls. The NSQIP defines the 30-day readmission window as a readmission occurring within 30 days of the index surgical procedure date. Readmissions to our hospital or any other hospital are included. Planned readmissions are also included in NSQIP’s registry. Principal readmission diagnosis and relationship between index operation and readmission are determined by the nurse reviewers (Table 1). Reviewers do not indicate if the readmission was considered potentially preventable at the time of presentation. University HealthSystem Consortium University HealthSystem Consortium (UHC) is an alliance of 116 academic medical centers and 275 affiliated hospitals that reports risk-adjusted performance metrics to its member institutions based on administrative data. The UHC defines the 30-day readmission window as a readmission occurring within 30 days of the index admission discharge date, similar to the proposed Medicare measure. The readmission is assigned a primary diagnosis code using the following algorithms: diagnosis-related group (Centers for Medicare and Medicaid Services [CMS]), all patient refined diagnosis-related group (3M), Clinical Classification System Category (Agency for Healthcare Research and Quality), and ICD-9 complication codes. If the readmission diagnosis code is any of the following it is considered planned and ultimately excluded: scheduled chemotherapy, radiation therapy, or dialysis treatment; same-day transfer to psychiatric facility, oncology ward, or inpatient rehabilitation; alcohol and
30-Day Readmission Criteria by Method
NSQIP criteria
Nurse reviewer medical record interpretation and/or patient interview
University HealthSystem Consortium Criteria
If any of the following criteria are met: Index hospitalization DRG ¼ readmission DRG Index hospitalization APR-DRG ¼ readmission APR-DRG Principal readmission diagnosis is a complication code (ICD-9 codes 996.00e999.9) Clinical classification system category of index hospitalization principal diagnosis ¼ clinical classification system category of readmission principal diagnosis Clinical classification system category of index hospitalization primary procedure ¼ clinical classification system category of readmission primary procedure
APR-DRG, all patient refined diagnosis-related group; DRG, diagnosis-related group.
1152
Hechenbleikner et al
Readmissions and Data Collection
drug treatment admission; and labor and delivery admission. Planned readmissions for any other reason are not excluded. Based on these 4 diagnosis code algorithms, a readmission is identified as related or unrelated to the index admission; all related readmissions are considered potentially preventable. A readmission is defined as related when certain index admission diagnosis codes match readmission diagnosis codes (Table 1). Surgeon case review All readmissions identified based on the 2 methodologies (NSQIP and UHC) were confirmed by physician chart review. The relationship to index operation or admission was determined clinically based on the patient’s presenting symptoms, laboratory data, and imaging studies. Readmissions were then evaluated for whether they were necessary at the time of the postoperative patient’s presentation. Readmissions were considered clinically necessary at the time of presentation if patients met one or more of the following criteria: ICU admission; unplanned return to the operating room; nonoperating room invasive procedure; placement of peripherally inserted central catheter; and abnormal vital signs, laboratory results, or radiographic data as defined in Table 2. Readmissions that did not meet any of these criteria were considered potentially unnecessary from a clinical standpoint. The primary readmission diagnosis or underlying complication was then assessed to determine whether it occurred during the index admission or after hospital discharge. Finally, if there was an underlying complication that led to the 30-day readmission, it was assessed for preventability based on clinical judgment to determine whether it could have been avoided during the index hospital stay and/or with enhanced outpatient care efforts. Statistical analysis A multivariate logistic regression analysis for risk factors predictive of readmission was performed using the R statistical programming language (R Foundation for Statistical Computing). Risk factors identified as contributing substantially to readmission were selected using a stepwise procedure.4 Patient-specific, intraoperative, and postoperative variables (Table 3) from the NSQIP database were included in the regression analysis. This study was approved by the Institutional Review Board at Johns Hopkins Hospital.
RESULTS We identified 735 patients who underwent colorectal operations between July 2009 and November 2011. Median age was 56 years (range 19 to 96 years). Six hundred and
J Am Coll Surg
Table 2. Criteria for Clinically Necessary 30-Day Readmission at the Time of the Patient’s Presentation At the time of rehospitalization, one or more of the following clinical criteria must be met: Major criteria ICU admission Unplanned return to the operating room Nonoperating room invasive procedure* Peripherally inserted central catheter (ie, for hydration, antibiotics, etc) Minor criteria Abnormal vital signs Temperature >38.3 C or <36 C Tachycardia (heart rate 110 bpm) Absolute hypotension (systolic blood pressure <90 mmHg) Clinical documentation of orthostatic hypotension Acute renal failure defined as 0.5 mg/dL increase in serum creatinine from a baseline of 1.9 mg/dL 1.0mg/dL increase in serum creatinine from a baseline of 2.0e4.9 mg/dL 1.5 mg/dL increase in serum creatinine from a baseline of 5.0 mg/dL White blood cell count >12,000 or <4,000 cells/mm3 Severe electrolyte imbalances or hypoglycemia requiring treatment Drop in hemoglobin count requiring blood transfusion CT scan evidence of bowel obstruction or anastomotic leak Bowel obstruction requiring nasogastric tube placement or parenteral nutrition *Percutaneous abdominal drain, endoscopy, bronchoscopy, cardiac catheterization, biliary drain placement, etc.
thirty-four (86%) patients underwent colon resections and 101 (14%) underwent rectal resections. Two hundred and ninety-four patients had ostomies created as part of their index operation (219 ileostomies, 75 colostomies). Two hundred and forty-one procedures (33%) were laparoscopic and 73 (10%) operations were done emergently. The most common indications for the procedures were 228 colorectal cancer (31%), 181 inflammatory bowel disease (24.6%), and 45 diverticulitis (6.1%; Table 3). The most common morbidity in the NSQIP database was surgical site infection (SSI) with 166 patients (22.5%) being diagnosed within 30 days of surgery; most SSIs were superficial incisional (116 patients [15.8%]) followed by organ space (44 patients [6%]). Significant risk factors for readmission in this group of patients were body mass index (odds ratio [OR] ¼ 1.05; p ¼ 0.001), presence of an ostomy (OR ¼ 2.59; p < 0.001), superficial incisional SSI (OR ¼ 1.74; p ¼ 0.043), deep incisional SSI (OR ¼ 14.78; p ¼ 0.004), and organ space SSI (11.54; p < 0.001). The 30-day unplanned readmission rate in NSQIP was 107 of 735 patients (14.6%) and in UHC was 129 of 735 patients (17.6%; Fig. 1). Median time to readmission was
Hechenbleikner et al
Vol. 216, No. 6, June 2013
Table 3.
Readmissions and Data Collection
1153
Characteristics of Readmitted and Nonreadmitted Patients
Characteristics
Patient-specific* Age, y, mean (range) Male sex, n (%) BMI, mean (range) Diabetes mellitus, n (%) Current smoker, n (%) Steroid use, n (%) Albumin, g/dL, mean (range) Emergent case, n (%) Surgical indication, n (%) Colorectal cancer (n ¼ 228) Diverticulitis (n ¼ 45) Inflammatory bowel disease (n ¼ 181) Other cancer (n ¼ 32) Miscellaneous (n ¼ 249)y Intraoperative, n (%) Type of resection Colon (n ¼ 634) Rectal (n ¼ 101) Type of approach Open (n ¼ 494) Laparoscopic (n ¼ 241) Presence of ostomy, n (%) Yes No Length of operation, h, mean (range) Postoperative, n (%) SSI Superficial incisional SSI Deep incisional SSI Organ space SSI Length of stay, d, median (range) 30-d mortality
Readmitted patients (n ¼ 129)
57 67 29 22 24 23 3.8 15 39/228 16/45 31/181 9/32 34/249
(22e86) (52) (15e59) (17) (18.6) (17.8) (1.7e5.2) (11.6) (17) (35.6) (17) (28) (14)
Nonreadmitted patients (n ¼ 606)
56 285 27 50 98 91 3.9 58 189/228 29/45 150/181 23/32 215/249
(19e96) (47) (13e58) (8.3) (16.2) (15) (1.1e5.1) (9.6) (83) (64.4) (83) (72) (86)
p Value
Odds ratio
NS NS 0.001 NS NS NS NS NS
NS NS 1.05 NS NS NS NS NS
0.002 NS 0.001 NS <0.001
0.35 NS 0.32 NS 0.24
108/634 (17) 21/101 (21)
526/634 (83) 80/101 (79)
NS NS
NS NS
96/494 (19) 33/241 (14)
398/494 (81) 208/241 (86)
NS NS
NS NS
74 (57) 55 (43) 4.9 (1.5e12.2)
220 (36) 386 (64) 4.2 (1e14)
<0.001 NS NS
2.59 NS NS
58 27 4 27 9 1
108 89 2 17 7 19
<0.001 0.043 0.004 <0.001 NS NS
3.39 1.74 14.78 11.54 NS NS
(45) (21) (3.1) (21) (3e65) (0.8)
(18) (14.7) (0.3) (2.8) (1e153) (3.1)
*All data for patient-specific and intraoperative variables were collected at the time of the index operation. y Miscellaneous includes multiple indications, such as volvulus, rectal prolapse, benign neoplasm, vascular insufficiency, accidental perforation, diverticulosis, etc. BMI, body mass index (calculated as kg/m2); SSI, surgical site infection.
8 days (range 1 to 30 days) and median length of hospitalization after readmission was 5 days (range 1 to 53 days). The NSQIP registry included 1 patient with a planned readmission for a right hemicolectomy. The UHC readmission cohort included 3 patients with planned readmissions (portal vein embolization, ostomy reversal, and right hemicolectomy). All planned readmissions were excluded from our analysis but were ultimately reported by NSQIP and UHC, respectively. Although the majority of unplanned readmissions (98 patients) were identified in both methodologies, multiple discrepancies were noted. The NSQIP method identified 9 readmissions not found with the UHC method because the
readmission occurred at another hospital (7 patients) or met UHC exclusion criteria (1 patient was a psychiatric facility transfer; 1 patient was an inpatient rehabilitation transfer). The UHC method identified 31 readmissions not found in NSQIP; 20 were because of the discrepancy in the 30-day readmission window definition (UHC defines the 30-day window as beginning at the date of discharge as opposed to the date of index operation) and 11 because they were missed by clinical reviewers. Next, we examined the diagnosis associated with readmission and whether the readmission was related to the index operation or admission based on NSQIP (nurse reviewer determination) and UHC criteria (diagnosis
1154
Hechenbleikner et al
Readmissions and Data Collection
J Am Coll Surg
Figure 1. Thirty-day readmission rates as determined by NSQIP and University HealthSystem Consortium (UHC) methods.
code algorithm), respectively. The NSQIP reviewers evaluate the relationship between a readmission and the index operation but do not currently indicate whether the readmission was considered preventable. Based on UHC methodology, if the readmission is related to the index admission, the readmission is deemed potentially preventable. These determinations were compared with surgeon chart review to assess accuracy. The principal diagnoses or complications that led to the readmission were divided into the following categories: bleeding, dehydration, infection, gastrointestinal related, and other (Table 4). In the NSQIP registry, the percentage of readmissions related to index surgical procedure was 72%. In the UHC registry, the percentage of readmissions related to index hospitalization was 51%. Surgeon case review demonstrated higher rates of readmissions related to index operation or admission at 83% and 86%, respectively (Table 4); the majority of readmissions were due to postoperative complications. The most frequent complications requiring readmission were SSI (46 of 129 patients [36%]), dehydration with high ileostomy output (20 of 129 patients [15.5%]), and gastrointestinal issues, including partial bowel obstruction or paralytic ileus (17 of 129 patients [13%]). In general, UHC consistently reported lower rates of readmissions related to index hospital stay compared with other methods. For example, UHC reported that 10% of readmissions for dehydration were related to the index admission as compared with surgeon case review at 97% (Table 4).
Identification of readmissions for infection as related to index operation or admission was most consistent in the 2 methodologies, 93% NSQIP vs 84% UHC, respectively. In addition, criteria were established to determine if, at the time the patient presented for readmission, this admission was clinically necessary for safe patient care (Table 2). By surgeon review, we established that 17 of 129 (13%) readmissions were potentially preventable from a clinical perspective; none of these patients met either major or minor criteria summarized in Table 2. The clinically avoidable readmissions were in the following categories: superficial SSI (5 patients), mild dehydration (4 patients), gastrointestinal-related issues (5 patients), and 3 other patients (skin excoriation, kidney stones, and failure to thrive). One hundred and 12 readmissions (87%) were determined to be clinically necessary at the time of presentation; 61 patients (54%) met one or more major criteria, 51 patients (46%) met one or more minor criteria alone (Table 5). At the time of presentation, 26 of 30 (87%) readmissions for complications of dehydration were not avoidable. The majority (22 of 26 [85%]) of these patients, had an ostomy (20 ileostomy, 2 colostomy). Of the 46 readmissions for SSI, 41 (89%) were necessary for safe patient care; many of these patients had abnormal vital signs or significant leukocytosis at time of presentation or required invasive procedures, ICU admissions, and/or surgical exploration during readmission. Eleven patients required ICU admission for severe dehydration (n ¼ 5) or organ
Vol. 216, No. 6, June 2013
Hechenbleikner et al
Readmissions and Data Collection
1155
Table 4. Accuracy of 30-Day Readmissions Related to Index Operation or Admission in NSQIP and University HealthCare System
Patient groups
NSQIP All patients Principal readmission diagnosis categoryy Bleeding Dehydration Infection Gastrointestinal Other University HealthCare System All patients Principal readmission diagnosis category Bleeding Dehydration Infection Gastrointestinal Other
Total no. of readmissions
Readmissions related to index operation n %
Surgeon case review n %
71*
51
72
59
83
3 16 27 16 9
0 9 25 11 6
0 56 93 69 67
2 15 25 13 4
67 94 93 81 44
129
66
51
111
86
4 30 51 27 17
1 3 43 14 5
25 10 84 52 29
3 29 46 25 8
75 97 90 93 47
*NSQIP began tracking readmission relatedness to index operation in spring 2010; only 71 patients were reviewed. y Readmission categories: bleeding (ie, gastrointestinal hemorrhage); other (ie, failure to thrive, medication overdose); gastrointestinal (ie, ileus, bowel obstruction); dehydration (ie, high ostomy output); and infection (urinary tract infection, surgical site infection).
space SSI (n ¼ 6) and 8 patients required reoperation for organ space SSI (n ¼ 5) or deep incisional SSI (n ¼ 3). Overall, 39 patients required percutaneous interventional and/or endoscopic procedures; 20 SSI patients (19 organ space, 1 deep incisional) required percutaneous abdominal drainage. Among all 129 patients who were readmitted, in 15 patients (11.6%) a complication was diagnosed during the index admission compared with 114 patients (88.4%) who were diagnosed with a complication at the time of readmission (data not shown). Complications diagnosed during the index admission were dehydration in 4 patients; gastrointestinal-related in 3 patients; SSIs in 7 patients; and deep venous thrombosis in 1 patient. Complications diagnosed at readmission were bleeding in 4 patients; dehydration in 26 patients; gastrointestinal-related in 24 patients; pneumonia in 1 patient; diverticulitis in 1 patient; colitis in 1 patient; cellulitis in 1 patient; SSI in 39 patients; urinary tract infection in 1 patient; and other (ie, myocardial infarction, kidney stones) in 16 patients. Among the 15 patients diagnosed with complications at the time of index admission, 10 of the patients (67%) had complications that could have been potentially avoided with different treatment regimens. Readmissions for dehydration with high ostomy output (n ¼ 4) might have been avoided with enhanced patient education, medication reconciliation, and/or outpatient care coordination. Organ space SSIs with anastomotic leak (n ¼ 3) requiring readmission might have been prevented if an
ostomy had been created during the index operation. One organ space SSI (abscess, no anastomotic leak) readmission might have been prevented with better inpatient care; this patient had a persistent intra-abdominal fluid collection after percutaneous abdominal drainage, which could have been drained again before discharge. Two patients with superficial SSIs were readmitted 1 day after hospital discharge. In this case, both improved inpatient efforts, including aggressive local wound care, and enhanced care transitions might have prevented these readmissions. Among the 114 complications diagnosed at readmission, 31 (27%) might not have required rehospitalization with enhanced patient-care efforts. Twentytwo dehydration readmissions (16 high ostomy output, 3 diarrhea, 1 vomiting, 1 poor fluid intake, and 1 misuse of diuretics), 5 gastrointestinal-related readmissions (1 severe constipation, 1 leakage around ostomy, 2 diarrhea, and 1 colostomy retraction), and 4 other readmissions (2 skin excoriation, 1 hyponatremia, and 1 misuse of narcotics) might have been prevented with improved outpatient care coordination. Overall, readmissions for SSIs and other infectious complications diagnosed after hospital discharge were very difficult to assess for preventability because of the sparse documentation in the medical record between index admission and rehospitalization.
DISCUSSION Hospital readmissions are emerging as a leading quality indicator and prevention has been targeted as a strategy
1156
Table 5.
Hechenbleikner et al
J Am Coll Surg
Readmissions and Data Collection
Surgeon Case Review of 30-Day Readmissions
Clinically necessary vs unnecessary 30-d readmissions
Bleeding (n ¼ 4 [3%]) n %
Necessary readmissions Major criteria ICU admission Unplanned return to operating room Nonoperating room invasive procedure Peripherally inserted central catheter Minor criteria (alone) Potentially unnecessary readmissions
4 3 2 0 3 0 1 0
100 75 50 75 25
Dehydration (n ¼ 30 [23%]) n %
26 11 5 1 3 5 15 4
to improve patient safety and reduce health care expenditures. Beginning in October 2012, CMS started paying hospitals 1% less for excess readmissions for patients with 3 common diagnoses (ie, acute myocardial infarction, heart failure, pneumonia). The same provision will increase the penalty to 2% in September 2013 and 3% in September 2014. This rule is expected to be extended to a hospital-wide all-cause unplanned readmission measure in 2013 that will include surgical patients. The CMS comprehensive all-cause, hospital-wide readmission measure will identify patients with unplanned readmissions within 30 days of discharge.5 To set benchmarks for unplanned, related surgical readmissions, the hospital-wide readmission measure will use algorithms matching procedure codes and discharge diagnosis categories from administrative claims data, similar to UHC methodology. Our study of readmission definitions and principal readmission diagnoses using administrative data and NSQIP as compared with physician chart review revealed multiple discrepancies. Although administrative data identified an unplanned readmission rate of 17.6%, NSQIP found the rate to be 14.6%. More importantly, we found that administrative data were not able to consistently determine whether a readmission was related to the index hospitalization. Although physician case review determined most readmissions were related to postoperative complications, administrative data frequently failed to relate the unplanned readmissions to the index admission or identify a postoperative complication as the reason for readmission. The accuracy varied based on the postoperative complication. Administrative data were most accurate at relating postoperative infections like SSI and least accurate at relating dehydration to the index hospitalization. We postulate that this is because postoperative infection is frequently coded with an ICD-9 complication code and this is automatically identified as related to index admission; in contrast, dehydration, the second
87 42.3 19.2 3.8 11.5 19.2 57.7 13
Infection (n ¼ 51 [40%]) n %
46 32 6 8 21 6 14 5
90 70 13 17.4 45.7 13 30 10
Gastrointestinal (n ¼ 27 [21%]) n %
22 8 1 2 5 3 14 5
81.5 36.4 4.5 9.1 22.7 13.6 63.6 18.5
Other (n ¼ 17 [13%]) n %
14 7 2 1 5 1 7 3
82 50 14.3 7.1 35.7 7.1 50 18
Totals (n ¼ 129) n %
112 61 16 12 37 15 51 17
87 54 14.2 10.7 33 13.4 46 13
leading cause of readmissions in this population, is not. Depending on the type of surgical patient (colorectal vs vascular vs spinal), administrative data could markedly vary in their ability to establish the relationship between index admission and readmission but current models do not factor in case-mix adjustment. Additional refinement of these algorithms is needed before administrative data can be considered for establishing benchmarks that will impact financial reimbursement policies. Emerging evidence suggests that surgical complications are an important factor in postoperative hospital readmissions.6 Studies of general surgery patients have suggested that readmissions in this population are often the result of operative complications, including SSI and bowel obstruction, as opposed to an exacerbation of underlying medical conditions.7-9 In surgical patients with postoperative complications, readmission is frequently clinically necessary for safe care. In our study, administrative data (UHC) predominantly defined readmissions as potentially preventable by identifying the principal readmission diagnosis as a complication code. The majority of preventable readmissions identified by administrative data in our study were related to postoperative infections. Based on clinical judgment, we determined that most of these readmissions were necessary for safe patient care at the time of presentation; 45.7% and 17.4% of patients required invasive procedures or reoperation, respectively (Table 5). Additional study is needed to determine if SSIs are preventable in colorectal surgery. We did, however, identify several postoperative complications (organ space and superficial SSIs) diagnosed during index admission that might have not required rehospitalization with enhanced inpatient treatment. Unfortunately, all-cause readmission policies like those adopted by CMS might have the perverse effect of incentivizing physicians not to accept patients at high risk for readmission or not to readmit postoperative patients when medically indicated. In other areas, implementation of hospital-based pay-for-performance programs and
Vol. 216, No. 6, June 2013
Hechenbleikner et al
benchmarks has demonstrated conflicting evidence to date for improving patient outcomes such as mortality.10,11 In colorectal surgery, prevention of readmission for an acutely ill postoperative patient could delay appropriate care and potentially be harmful. Accordingly, resources allocated toward preventing surgical readmissions should be redirected toward individual and systems-based practices for preventing complications. We previously12 reported on the implementation of an interdisciplinary program (CUSP [Comprehensive Unit-Based Safety Program]) to decrease SSI rates in a high-risk surgical cohort by standardizing preoperative and intraoperative techniques. Hospital investment in inpatient quality-improvement programs might be the most effective strategy for preventing surgical readmissions. We did identify, by physician chart review, a subset of postoperative complications that might not have required readmission with enhanced outpatient care coordination. The majority of these readmissions were for patients with dehydration and high ostomy output but also included constipation, skin excoriation, and misuse of narcotics. Although no care-coordination interventions have focused on surgical patients, there are reports of effective models in medical patients. Most studies have focused on transitional care, such as discharge planning, medication reconciliation, and timely outpatient follow-up. In a 2011 systematic review of 43 studies, the most common interventions tested were predischarge patient education, discharge planning, and/or postdischarge telephone calls.13 Only 5 randomized trials showed a significant absolute risk reduction in 30-day readmissionsd1 with a single intervention and 4 with multiple or “bundled” interventions. The authors concluded that no single intervention or bundled interventions appeared to consistently decrease 30-day readmissions and that additional research is required. Centers have begun similar interventions in surgical patients with aggressive follow-up in the postdischarge postoperative period. These efforts include frequent calls to the patient and early office visits with a nurse or physician. More study is needed to evaluate the efficacy of such interventions in this high-risk population. There were several important limitations to this study. First, this was a retrospective cohort study with a small number of patients. Second, varying definitions used by NSQIP and UHC for determining 30-day readmissions and the relationship to index admission made qualitative comparisons difficult. Third, there are no established guidelines in the literature for determining preventable or clinically unnecessary hospital readmissions in surgical patients. Finally, our analysis is based on a select, high-risk surgical cohort at a single tertiary referral academic center and might not be representative of readmissions at other institutions.
Readmissions and Data Collection
1157
CONCLUSIONS Financial penalties placed on hospitals for excess readmissions are imminent and are currently based on administrative data. All-cause readmission benchmarks that are derived from administrative data should be used cautiously in surgical patients until additional validation and research has been performed. Within a select surgical cohort, we demonstrate that the ability to relate index and readmitting conditions and clinically unnecessary readmissions is more reliable when using surgeon-endorsed definitions and an independent clinical reviewer. In our sample, SSI was the only postoperative complication that was consistently linked back to the index admission and the majority of these readmissions were necessary for safe patient care, based on physician chart review. Frequently, these patients required invasive procedures or reoperation, so any delay of treatment could have resulted in considerable harm. With the addition of well-defined, consistent criteria, prospective clinical registries such as NSQIP that use independent clinical reviewers might be able to track postoperative complications driving readmissions and ultimately determine which interventions can help reduce readmissions. If most surgical readmissions are due to complications that require treatment and hospitalization at the time of presentation, the best way to decrease readmissions is to prevent complications from ever happening. Future research and resources should focus on perioperative interventions to decrease surgical complications and how these interventions compare with other efforts, such as outpatient care coordination, to decrease 30-day readmissions. Author Contributions Study conception and design: Hechenbleikner, Makary, Gearhart, Efron, Wick Acquisition of data: Hechenbleikner, Bennett, Wick Analysis and interpretation of data: Hechenbleikner, Samarov, Wick Drafting of manuscript: Hechenbleikner, Makary, Samarov, Bennett, Gearhart, Efron, Wick Critical revision: Hechenbleikner, Makary, Samarov, Bennett, Gearhart, Efron, Wick Acknowledgment: The authors thank Lucy Mitchell, RN, MA, QI Specialist, and SCNR ACS NSQIP, for assistance with acquisition, analysis, and interpretation of data as well as Jason Miller, Director of Clinical Analytics, for assistance with UHC data. REFERENCES 1. Medicare Payment Advisory Commission. Report to Congress: promoting efficiency in Medicare, Chapter 5. Available at: www.medpac.gov. Accessed June 15, 2007.
1158
Hechenbleikner et al
Readmissions and Data Collection
2. Rau J. Medicare to penalize 2,211 hospitals for excess readmissions. Kaiser Health News. Available at: www.kaiserhealthnews. org. Accessed August 13, 2012. 3. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med 2009;360:1418e1428. 4. Hastie TJ, Pregibon D. Generalized linear models. In: Chambers JM, Hastie TJ, eds. Statistical Models in S. Pacific Grove, CA: Wadsworth & Brooks/Cole; 1991:195e248. 5. Horwitz L, Partovian C, Lin Z, et al. Hospital-wide (all-condition) 30-day risk-standardized readmission measure, draft measure methodology report. Yale New Haven Health Services Corporation/Center for Outcomes Research and Evaluation (YNHHSC/CORE). Available at: www.cms.gov. Accessed Aug 10, 2011. 6. Kassin MT, Owen RM, Perez SD, et al. Risk factors for 30-day hospital readmission among general surgery patients. J Am Coll Surg 2012;215:322e330. 7. Reddy DM, Townsend CM Jr, Kuo YF, et al. Readmission after pancreatectomy for pancreatic cancer in Medicare
8.
9. 10. 11. 12. 13.
J Am Coll Surg
patients. J Gastrointest Surg 2009;13:1963e1974; discussion 1974e1975. Schneider EB, Hyder O, Brooke BS, et al. Patient readmission and mortality after colorectal surgery for colon cancer: impact of length of stay relative to other clinical factors. J Am Coll Surg 2012;214:390e398; discussion 398e399. Ozturk E, Kiran RP, Remzi F, Fazio VW. Early readmission after ileoanal pouch surgery. Dis Colon Rectum 2009;52:1848e1853. Jha AK, Joynt KE, Oray EJ, et al. The long-term effect of premier pay for performance on patient outcomes. N Engl J Med 2012;366:1606e1615. Sutton M, Nikolova S, Boaden R, et al. Reduced mortality with hospital pay for performance in England. N Engl J Med 2012;367:1821e1828. Wick EC, Hobson DB, Bennett JL, et al. Implementation of a surgical comprehensive unit-based safety program to reduce surgical site infections. J Am Coll Surg 2012;215:193e200. Hansen LO, Young RS, Hinami K, et al. Interventions to reduce 30-day rehospitalization: a systematic review. Ann Intern Med 2011;155:520e528.