Postoperative 30-Day Readmission: Time to Focus on What Happens Outside the Hospital

Melanie S. Morris, MD; Laura A. Graham, MPH; Joshua S. Richman, MD, PhD; Robert H. Hollis, MD; Caroline E. Jones, MD; Tyler Wahl, MD; Kamal M. F. Itani, MD; Hillary J. Mull, PhD; Amy K. Rosen, PhD; Laurel Copeland, PhD; Edith Burns, MD; Gordon Telford, MD; Jeffery Whittle, MD, MPH; Mark Wilson, MD; Sara J. Knight, PhD; Mary T. Hawn, MD, MPH

Disclosures

Annals of Surgery. 2016;264(4):621-631. 

In This Article

Results

Our study sample included 237,441 surgeries: 101,501 (42.7%) orthopedic, 93,019 (39.2%) general surgery, and 42,921 (18.1%) vascular surgery. Patients were mostly white (78.0%) elderly men. The overall 30-day unplanned postoperative readmission rate was 11.1% with significant differences among surgical specialties (vascular 15.4%, general 12.9%, and orthopedic 7.6% P < 0.001). Readmission rates decreased slightly across the 7 years of follow-up from 11.3% in 2008 to 10.3% in 2014. A total of 1994 (0.84%) patients died within 30 days of hospital discharge, of whom 31.5% had unplanned 30-day readmission before death. Patient-level variables and unadjusted rates of readmission are shown in Table 1. Patients who were widowed had partially or totally dependent functional status, and those who were admitted to the hospital from a nursing home were more likely to be readmitted. Racial differences in readmission were also observed; Asians had the highest readmission rates (13.0%) followed by blacks (12.0%) and whites (11.0%). Patients with a mental health diagnosis had higher readmission rates than the nonmental health cohort (schizophrenia: 16.1% vs 11.0%; depression: 12.5% vs 10.8%).

As shown in Table 1, patients with preoperative acute renal failure and those on dialysis had high rates of readmission, 23.8% and 26.0% respectively. Patients with cardiovascular disease also had high rates of readmission: these included diagnoses of congestive heart failure (23.5%), recent MI (21.3%), angina (17.1%), and previous cardiac surgery (15.3%). Patients with evidence of more than 1 healthcare encounter within 6 months before surgery were also more likely to be readmitted: ER visits 18.6%, P < 0.001; inpatient admissions 19.0%, P < 0.001).

Next, we examined operative characteristics and hospital course to assess their relationship with unadjusted readmission rates (Table 2). Readmitted patients had longer operative times, were more likely to be emergent cases, and underwent higher work relative value unit (RVU) procedures than the overall surgical cohort. Procedure type was strongly associated with readmission rates. Severity of wound classification was also associated with increased readmission rates: clean 9.0%, clean/contaminated 13.6%, contaminated 15.8%, and infected 17.1% (P < 0.001). Higher readmission rates were related to whether postoperative complications occurred before hospital discharge or not: no complication (10.6%), 1 complication (18.3%), and more than 1 complication (19.6%, P < 0.001). Discharge destination other than home was associated with an increased readmission rate (14.7%).

Patient-level postoperative vital signs, pain scores, and laboratory values were associated with rates of readmission (Table 2). Maximum postoperative pain scores (categorized as <8, 8, 9, or 10) were associated with escalating readmission rates: 9.6%, 10.7%, 11.6%, and 13.2% respectively (P < 0.001).

Figure 1 shows the daily hazard for each readmission reason along with the distribution of all readmissions across time since hospital discharge, stratified by specialty. The most common reasons for readmission were wound complications (30.7%) followed by gastrointestinal (GI) reasons (eg, ileus or obstruction [16.1%]), bleeding (4.9%), and fluid/electrolyte abnormalities/genitourinary (7.5%). Wound complications peaked early after discharge for general surgery patients, but continued to increase for vascular patients up to 2 weeks after discharge before declining. Wound complications for orthopedic patients remained low throughout the 30 days postdischarge. GI reasons were common for general surgical readmission within the first week postdischarge, but were uncommon in vascular or orthopedic patients. General surgery readmissions were most common within the first week, but then decreased below those of vascular readmissions. Orthopedic patients had the lowest risk of readmission at all timepoints. For all patients, the risk of readmission was highest within the first week, and then tapered off over time, although readmission remained higher for vascular patients. Figure 2 shows the most common readmission reasons stratified by specialty.

Figure 1.

Instantaneous hazard and distribution of readmission across time since Index Hospital Discharge.

Figure 2.

Reasons for readmission stratified by specialty.

The risk-adjusted model of unplanned readmission to examine the contributions of patient-level preoperative variables, operative characteristics, and postoperative predischarge variables to explain variation in predicting readmission is depicted in Table 3. Procedure Current Procedure Terminology codes, facility, laboratory values, prior healthcare utilization, postoperative pain scores, and ASA class contributed the most to predicting readmission. Using information available at the time of hospital discharge, models were only able to explain 10.3% of the variability in readmission risk. Preoperative patient-level factors contributed the most to the predictive model (R 2 7.0%, c-statistic 0.67) with small increases in predictive ability with inclusion of operative variables (R 2 9.0%, c-statistic 0.69) and postoperative variables (R 2 10.4%, c-statistic 0.71). The models were better at explaining overall variability in readmission risk in vascular patients than in orthopedic or general surgery patients (R 2 9.6%, 9.0%, and 8.3%, respectively).

The individual readmission risk models developed to evaluate our ability to predict specific reasons for readmission were stratified by specialty (Fig. 3). Wound complications were the most common readmission reason, but they performed poorly in predicting risk of readmission for orthopedic and general surgery patients (R 2 5.4% and 6.1%). Our best predictive model was for orthopedic patients readmitted with pneumonia; however, even here, we could explain only 14% of the variation in readmission risk at the time of discharge.

Figure 3.

Model R-square for specific reasons of readmission stratified by specialty.

To explore the prediction of readmission using all available information, we constructed overall models that also included postdischarge complications. Adding this information to factors available at discharge increased our predictive ability considerably, allowing us to explain 19.6% of readmissions (model R 2 19.6%, c-statistic 0.76).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....