Background
Hospital readmissions are costly, frequent, and increasingly under public scrutiny. With increased financial constraints on the medical environment, understanding the drivers of unscheduled readmissions following gynecologic surgery will become increasingly important to value-driven care.
Objective
The current study was conducted to identify risk factors for 30-day readmission following hysterectomy for benign and malignant indications.
Study Design
A retrospective cohort study was conducted from 2008 through 2010 of all nongravid hysterectomies at a single tertiary care academic medical center. Clinical, perioperative, and physician characteristics were collected. Multivariable logistic regression models were used to identify predictors of 30-day readmission, stratified by malignant and benign indications for hysterectomy.
Results
Among 1649 women who underwent a hysterectomy (1009 for benign indications and 640 for malignancy), 6% were subsequently readmitted within 30 days (8.9% for malignancy vs 4.2% for benign; P < .0001). The mean time to readmission was 13 days (15 days for malignancy vs 10 days for benign; P = .004). The most common reasons for readmission were gastrointestinal (38%) and infectious (34%) etiologies, and 11.6% of readmitted patients experienced a perioperative complication. Among women undergoing hysterectomy for benign indications, a history of a laparotomy, including cesarean delivery (adjusted odds ratio [AOR], 2.12; 95% confidence interval [CI], 1.06–4.25; P = .03), as well as a perioperative complication (AOR, 2.41; 95% CI, 1.00–6.04; P = .05) were both associated with a >2-fold increased odds of readmission. Among women undergoing hysterectomy for malignancy, an American Society of Anesthesiologists Physical Status Classification of III or IV (AOR, 1.92; 95% CI, 1.05–3.50; P = .03), a longer length of initial hospitalization (3 days AOR, 7.83; 95% CI, 1.33–45.99; P = .02), and an estimated blood loss >500 mL (AOR, 3.29; 95% CI, 1.28–8.45; P = .01) were associated with a higher odds of readmission; however, women who underwent a laparoscopic hysterectomy (AOR, 0.32; 95% CI, 0.12–0.86; P = .02) and who were discharged on postoperative day 1 (AOR, 0.16; 95% CI, 0.03–0.82; P = .02) were at a decreased risk of readmission. Physician and operative characteristics were not significant predictors of readmission.
Conclusion
This study found that malignancy, perioperative complications, and prior open abdominal surgery, including cesarean delivery, are significant risk factors for consequent 30-day readmission following index hysterectomy. It may be possible to identify patients at highest risk for readmission at the time of hysterectomy, which can assist in developing interventions to reduce such events.
Introduction
Hospital readmissions are frequent and costly, and are estimated to cost the US health care system >$12-17 billion annually. The Medicare Payment Advisory Committee estimated that 12.3% of readmissions are potentially preventable. Recent data suggest that nearly 1 in every 7 patients hospitalized for a major surgical procedure may be readmitted to the hospital within 30 days of discharge. There has been increasing recognition that surgical readmissions may arise from complications secondary to the care delivered at the time of the index procedure. Hence, policymakers have focused on unplanned postsurgical hospital readmissions as an objective metric for quality of care. Since 2009, the Centers for Medicare and Medicaid Services (CMS) has published 30-day readmission data for selected high-volume medical conditions. In 2010, as part of the Patient Protection and Affordable Care Act, CMS began to financially penalize hospitals with excessive readmissions. It is anticipated that readmission following elective, nonemergent procedures will soon be added to the public reporting and pay for performance.
Hysterectomy is the second most commonly performed surgical procedure worldwide, with >600,000 cases each year in the United States alone, following cesarean delivery. By age 60 years, nearly 1 in every 3 US women will have undergone a hysterectomy. However, data focusing on perioperative complications and readmission following hysterectomy remain relatively limited, and previously published readmission rates after hysterectomy have ranged from 3.4-12.0% based on the type of patient population. Two recently published large-scale national observational studies using data from the National Surgical Quality Improvement Program across several hundred hospitals provided benchmark data for the overall frequency of 30-day readmission following hysterectomy (3.4% and 3.9%) ; however, earlier studies have been limited due to the lack of available detailed patient-level data assessing clinical risk factors for readmission.
The current study was conducted to identify risk factors for 30-day readmission following hysterectomy for benign and malignant indications. Given the increasing interest surrounding postsurgical readmission as a patient quality and payment metric, further detailed data assessing relevant patient, clinical, intraoperative and postoperative, and physician characteristics associated with readmission following hysterectomy are timely and can inform further quality of care interventions.
Materials and Methods
Study setting and participants
This retrospective cohort study was conducted over a period of 36 months from Jan. 1, 2008, through Dec. 31, 2010. All women who underwent hysterectomies by the same academic faculty practice of generalist and subspecialist gynecologists at Massachusetts General Hospital (Boston, MA) were included in this analysis. Women who underwent a gravid hysterectomy at the time of cesarean delivery were excluded. Data were abstracted from electronic surgical and anesthesia medical records. This study was approved by the Partners Healthcare Institutional Review Board.
Study outcomes and covariates
The primary study outcome was an unplanned readmission within 30 days of hysterectomy. For each patient, clinical data were abstracted to characterize preoperative, intraoperative, and postoperative variables that could be identified with readmission following hysterectomy. Preoperative variables included: patient age; American Society of Anesthesiologists (ASA) Physical Status Classification; clinical indication for operation; body mass index (BMI); preoperative laboratory results; previous abdominal surgery (laparoscopy and laparotomy); surgeon years of experience; and yearly volume of cases per surgeon. History of open abdominal surgery included cesarean delivery as well as exploratory laparotomy. Number of years in gynecologic practice postresidency training was used as a proxy for surgeon experience. Hysterectomy volume (ie, average hysterectomies per year over the 3-year period) was assessed, with the distinction between high- and low-volume surgeons defined as ≥10 vs <10 hysterectomies per year.
Intraoperative variables were taken from electronic anesthesia records and included: operative method; anesthetic used; arterial line placement; operative duration; intraoperative vitals; estimated blood loss (EBL); volume of blood products administered; and unplanned intraoperative consultation. Length of surgery was defined as incision to skin closure time (minutes). Using intraoperative variables, a surgical Apgar score, which has been previously validated for use in both general and gynecologic surgery, was calculated based on EBL, lowest mean arterial pressure, and lowest heart rate. A prior analysis from this data set validated a surgical Apgar score in predicting perioperative morbidity among only women undergoing hysterectomy for malignant indications. To assess the overall extensiveness of each surgery–how radical it was–for our statistical models, we created a radicality score for each patient, which is consistent with earlier studies. A patient was assigned 1 point for each of the following procedures performed at the time of their surgery: small bowel resection, large bowel resection, splenectomy, cholecystectomy, and lymph node dissection.
Postoperative variables included: presence of malignancy on final pathology and primary site of malignancy, if applicable; uterine weight; postoperative length of stay; and readmission within 30 days of operation. Of note, borderline tumors were considered benign. For this reason, some patients with benign pathology underwent procedures typically associated with gynecologic oncology surgery (eg, lymphadenectomy or omentectomy). Perioperative complications were defined as follows: unplanned intensive care unit (ICU) admission; unplanned return to the operating room (within 30 days of hysterectomy); length of hospital stay >15 days; EBL >2 L (per anesthesia record); transfusion >4 U of packed red blood cells (PRBCs) within 72 hours of hysterectomy; vascular, nerve or visceral injury; sepsis; anastomotic leak; fistula; postoperative bowel obstruction; perioperative pulmonary embolus or deep vein thrombosis; myocardial infarction; stroke; and death within 30 days of operation. Reasons for 30-day readmission were organized based on organ system into the following categories: neurologic, cardiovascular, pulmonary, gastrointestinal, genitourinary, infectious, hematologic, and other.
Statistical analysis
Given the differences in clinical and operative characteristics by indication, data are presented separately for hysterectomies performed for benign compared to malignant indications. Patient, clinical, intraoperative and postoperative, and physician characteristics were compared between women undergoing hysterectomy for benign compared to malignant indications. Logistic regression models were utilized to examine the relationship between the primary outcome of readmission within 30 days of operation and each preoperative, intraoperative, and postoperative variable. In addition to using stepwise regression models, we identified confounding variables for 30-day readmission based on statistical significance in the univariate analysis and an a priori review of prior publications of readmission following hysterectomy. In multivariable analysis, we adjusted for the following variables: age, ASA class, prior abdominal surgery, operative approach, and perioperative complications. Additionally for patients with malignancy we adjusted for radicality of surgery. We present multivariable logistic regression models for predictors of 30-day readmission separately for women undergoing hysterectomy for benign compared to malignant indications with an adjusted odds ratio (AOR) and the corresponding 95% confidence interval (CI) and P value. We also present Kaplan-Meier survival curves for time to readmission. All statistical analysis was conducted with software (STATA, Version 10.0; StataCorp LP, College Station, TX). A P value < .05 was considered significant.
Materials and Methods
Study setting and participants
This retrospective cohort study was conducted over a period of 36 months from Jan. 1, 2008, through Dec. 31, 2010. All women who underwent hysterectomies by the same academic faculty practice of generalist and subspecialist gynecologists at Massachusetts General Hospital (Boston, MA) were included in this analysis. Women who underwent a gravid hysterectomy at the time of cesarean delivery were excluded. Data were abstracted from electronic surgical and anesthesia medical records. This study was approved by the Partners Healthcare Institutional Review Board.
Study outcomes and covariates
The primary study outcome was an unplanned readmission within 30 days of hysterectomy. For each patient, clinical data were abstracted to characterize preoperative, intraoperative, and postoperative variables that could be identified with readmission following hysterectomy. Preoperative variables included: patient age; American Society of Anesthesiologists (ASA) Physical Status Classification; clinical indication for operation; body mass index (BMI); preoperative laboratory results; previous abdominal surgery (laparoscopy and laparotomy); surgeon years of experience; and yearly volume of cases per surgeon. History of open abdominal surgery included cesarean delivery as well as exploratory laparotomy. Number of years in gynecologic practice postresidency training was used as a proxy for surgeon experience. Hysterectomy volume (ie, average hysterectomies per year over the 3-year period) was assessed, with the distinction between high- and low-volume surgeons defined as ≥10 vs <10 hysterectomies per year.
Intraoperative variables were taken from electronic anesthesia records and included: operative method; anesthetic used; arterial line placement; operative duration; intraoperative vitals; estimated blood loss (EBL); volume of blood products administered; and unplanned intraoperative consultation. Length of surgery was defined as incision to skin closure time (minutes). Using intraoperative variables, a surgical Apgar score, which has been previously validated for use in both general and gynecologic surgery, was calculated based on EBL, lowest mean arterial pressure, and lowest heart rate. A prior analysis from this data set validated a surgical Apgar score in predicting perioperative morbidity among only women undergoing hysterectomy for malignant indications. To assess the overall extensiveness of each surgery–how radical it was–for our statistical models, we created a radicality score for each patient, which is consistent with earlier studies. A patient was assigned 1 point for each of the following procedures performed at the time of their surgery: small bowel resection, large bowel resection, splenectomy, cholecystectomy, and lymph node dissection.
Postoperative variables included: presence of malignancy on final pathology and primary site of malignancy, if applicable; uterine weight; postoperative length of stay; and readmission within 30 days of operation. Of note, borderline tumors were considered benign. For this reason, some patients with benign pathology underwent procedures typically associated with gynecologic oncology surgery (eg, lymphadenectomy or omentectomy). Perioperative complications were defined as follows: unplanned intensive care unit (ICU) admission; unplanned return to the operating room (within 30 days of hysterectomy); length of hospital stay >15 days; EBL >2 L (per anesthesia record); transfusion >4 U of packed red blood cells (PRBCs) within 72 hours of hysterectomy; vascular, nerve or visceral injury; sepsis; anastomotic leak; fistula; postoperative bowel obstruction; perioperative pulmonary embolus or deep vein thrombosis; myocardial infarction; stroke; and death within 30 days of operation. Reasons for 30-day readmission were organized based on organ system into the following categories: neurologic, cardiovascular, pulmonary, gastrointestinal, genitourinary, infectious, hematologic, and other.
Statistical analysis
Given the differences in clinical and operative characteristics by indication, data are presented separately for hysterectomies performed for benign compared to malignant indications. Patient, clinical, intraoperative and postoperative, and physician characteristics were compared between women undergoing hysterectomy for benign compared to malignant indications. Logistic regression models were utilized to examine the relationship between the primary outcome of readmission within 30 days of operation and each preoperative, intraoperative, and postoperative variable. In addition to using stepwise regression models, we identified confounding variables for 30-day readmission based on statistical significance in the univariate analysis and an a priori review of prior publications of readmission following hysterectomy. In multivariable analysis, we adjusted for the following variables: age, ASA class, prior abdominal surgery, operative approach, and perioperative complications. Additionally for patients with malignancy we adjusted for radicality of surgery. We present multivariable logistic regression models for predictors of 30-day readmission separately for women undergoing hysterectomy for benign compared to malignant indications with an adjusted odds ratio (AOR) and the corresponding 95% confidence interval (CI) and P value. We also present Kaplan-Meier survival curves for time to readmission. All statistical analysis was conducted with software (STATA, Version 10.0; StataCorp LP, College Station, TX). A P value < .05 was considered significant.
Results
From Jan. 1, 2008, through Dec. 31, 2010, a total of 1649 women underwent a nongravid hysterectomy, of which 1009 (61%) were for benign indications and 640 (39%) for malignancy. A total of 100 women (6%) were readmitted within 30 days of operation. The readmission rate remained stable over time: in 2008, 32/501 (6.4%); in 2009, 34/567 (6.0%); and in 2010, 34/581 (5.9%) patients were readmitted. Women with a malignancy were >2 times more likely to be readmitted within 30 days of hysterectomy compared to women undergoing hysterectomy for benign indications (8.9% vs 4.2%; P < .0001).
Demographic, clinical, and operative characteristics
Tables 1 and 2 describe the clinical and operative characteristics overall, and separately between malignant compared to benign cases. The average age was 54.8 years, with an average BMI of 29.6 kg/m 2 . The most common operative approach was laparotomy (50.9%), followed by conventional laparoscopy (27.0%) and vaginal hysterectomy (17.5%). The most common indications for hysterectomy for benign indications included fibroid uterus (34.4%), menstrual disorders including endometriosis (30.7%), incontinence and/or prolapse (25.9%), and prophylactic or for precancerous lesions (23.2%). Among women undergoing hysterectomy for malignancy, the primary anatomic sites were endometrial (50%), ovarian (40%), and cervical (10%). Almost a third of women (32.1%) reported prior laparotomy, of whom close to half (47.8%) had undergone a prior cesarean delivery. There were 46 staff gynecologists who performed at least 1 hysterectomy during the study period. Of these, 6 were trained as gynecologic oncologists. Of those cases with malignancy on final pathology, 18 (2.8%) were primarily performed by nononcologists (generalists, reproductive surgery, or female pelvic medicine and reconstructive surgery specialists). Gynecologists on average had 13.7 mean years of surgical experience (since medical school graduation). One fifth (20.4%) of staff gynecologists performed >10 hysterectomies per year (ie, high volume), and these high-volume surgeons performed 78% of all hysterectomies. The mean length of surgery was 191 minutes, and the mean length of hospitalization was 3 days. Over a third (34.8%) of patients were discharged home on postoperative day 1.
| Variable | Overall, N = 1649 (%) | Malignancy, N = 640 (%) | Benign, N = 1009 (%) | ||||
|---|---|---|---|---|---|---|---|
| Readmission N = 57 (%) | No readmission N = 583 (%) | P value | Readmission N = 43 (%) | No readmission N = 966 (%) | P value | ||
| Age, mean (SD), y | 54.8 (12.2) | 61.1 (12.4) | 60.1 (11.9) | .55 | 49.2 (12.6) | 51.4 (11.0) | .20 |
| BMI, mean (SD), kg/m 2 | 29.6 (7.9) | 32.0 (9.2) | 30.9 (9.2) | .37 | 30.0 (6.9) | 28.7 (6.9) | .24 |
| Operative approach | |||||||
| Laparotomy | 840 (50.9) | 46 (80.7) | 368 (63.1) | .02 | 24 (55.8) | 402 (41.6) | .14 |
| Laparoscopy | 445 (27.0) | 5 (8.8) | 161 (27.6) | 10 (23.3) | 269 (27.8) | ||
| Vaginal | 288 (17.5) | 1 (1.8) | 7 (1.2) | 7 (16.3) | 273 (28.3) | ||
| Robotic | 76 (4.7) | 5 (8.8) | 47 (8.1) | 2 (4.7) | 22 (2.3) | ||
| ASA class | |||||||
| I | 148 (9.0) | 0 | 22 (3.7) | .01 | 6 (14.0) | 120 (12.5) | .03 |
| II | 1221 (74.0) | 32 (56.1) | 415 (71.1) | 26 (62.8) | 747 (77.4) | ||
| III | 272 (16.5) | 25 (43.9) | 142 (24.4) | 10 (23.3) | 95 (9.8) | ||
| IV | 7 (0.4) | 0 | 4 (0.7) | 0 | 3 (0.3) | ||
| Surgeon experience, y, median (IQR) | 16 (14–21) | 17 (15–19) | 17 (15–27) | .91 | 15 (12–21) | 15 (12–20) | .91 |
| Surgeon volume, >10 cases/y | 1301 (78.9) | 55 (96.5) | 558 (95.7) | .78 | 29 (67.4) | 659 (68.2) | .91 |
| Prior abdominal surgery a | |||||||
| None | 820 (49.7) | 26 (45.6) | 301 (51.6) | 15 (34.9) | 478 (49.5) | ||
| Laparoscopy | 299 (18.2) | 10 (17.5) | 98 (16.8) | .66 | 5 (11.6) | 186 (19.2) | <.01 |
| Laparotomy, including prior Cesarean delivery | 529 (32.1) | 21 (36.8) | 184 (31.6) | 23 (53.5) | 301 (31.2) | ||
| Prior cesarean delivery | 254 (15.4) | 7 (12.3) | 75 (12.9) | .89 | 14 (30.2) | 158 (16.3) | .01 |
| Indication for hysterectomy a | |||||||
| Pelvic mass | 328 (19.9) | 19 (33.3) | 132 (22.6) | .07 | 8 (18.6) | 169 (17.5) | .85 |
| Incontinence/prolapse | 275 (16.7) | 0 | 13 (2.2) | .25 | 7 (16.3) | 255 (26.4) | .13 |
| Uterine fibroids | 373 (22.6) | 0 | 25 (4.3) | .11 | 14 (33.0) | 334 (34.6) | .78 |
| Abnormal uterine bleeding | 316 (19.2) | 2 (3.5) | 9 (1.5) | .27 | 15 (34.9) | 290 (30.0) | .49 |
| Pelvic pain | 171 (10.4) | 0 | 17 (2.9) | .19 | 5 (11.6) | 149 (15.4) | .49 |
| Torsion | 2 (0.1) | 0 | 0 | – | 0 | 2 (0.3) | – |
| Endometritis | 3 (0.2) | 0 | 0 | – | 0 | 3 (0.4) | – |
| Prophylactic/risk reduction | 294 (17.8) | 3 (5.3) | 51 (8.7) | .36 | 14 (32.3) | 226 (23.4) | .16 |
| Baseline serum albumin, g/dL | |||||||
| ≤4 | 202 (20.8) | 11 (25.6) | 109 (26.8) | .85 | 1 (3.7) | 81 (17.4) | .07 |
| >4 | 770 (79.2) | 32 (74.4) | 297 (73.2) | 26 (96.3) | 415 (83.7) | ||
| Baseline hematocrit, % | |||||||
| ≤32 | 122 (7.5) | 4 (7.4) | 51 (8.9) | .75 | 1 (2.3) | 66 (6.9) | .07 |
| 32–42 | 1281 (78.7) | 46 (85.2) | 448 (77.8) | 40 (93.0) | 747 (78.4) | ||
| >42 | 225 (13.8) | 6 (11.1) | 77 (13.4) | 2 (4.7) | 140 (14.7) | ||
| Baseline white blood cell count, K/uL | |||||||
| ≤12,000 | 1547 (95.1) | 51 (91.1) | 541 (94.1) | .37 | 40 (93.0) | 915 (96.0) | .33 |
| >12,000 | 80 (4.9) | 5 (8.9) | 34 (5.9) | 3 (7.0) | 38 (4.0) | ||
| Baseline platelet count, K/uL | |||||||
| ≤150,000 | 28 (1.7) | 1 (1.8) | 8 (1.4) | .93 | 0 (0.0) | 19 (2.0) | .34 |
| 150,000–400,000 | 1386 (85.2) | 45 (80.4) | 472 (82.2) | 36 (83.7) | 833 (87.4) | ||
| >400,000 | 212 (13.0) | 10 (17.9) | 94 (16.4) | 7 (16.3) | 110 (10.6) | ||
| Baseline creatinine, mean (SD), mg/dL | .95 (2.07) | .92 (.19) | .91 (.3) | .80 | .88 (.18) | .99 (2.7) | .81 |
| Random serum glucose, mean (SD), mg/dL | 102.3 (35.1) | 111.3 (41.7) | 106.5 (39.0) | .38 | 96.4 (24.0) | 99.0 (31.6) | .61 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
