 |
|
||||
|
|||||
| About | Latest Articles | Back-Issues | Articles
|
Search | Instructions | Online Submission | Subscribe | Etcetera | Contact |
| |||||||||||||||||||||
|
Temporal variability of readmission determinants in postoperative vascular surgery patients MJ Lin1, F Baky2, BC Housley2, N Kelly2, E Pletcher3, JD Balshi1, SP Stawicki4, DC Evans21 Department of Surgery, St. Luke's University Health Network, Bethlehem, PA, USA 2 Department of Surgery, The Ohio State University College of Medicine, Columbus, OH, USA 3 Temple University School of Medicine – St. Luke's University Hospital Campus, Bethlehem, PA, USA 4 Department of Surgery; Department of Research and Innovation, St. Luke's University Health Network, Bethlehem, PA, USA
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0022-3859.188548
Introduction: Clinical information continues to be limited regarding changes in the temporal risk profile for readmissions during the initial postoperative year in vascular surgery patients. We set out to describe the associations between demographics, clinical outcomes, comorbidity indices, and hospital readmissions in a sample of patients undergoing common extremity revascularization or dialysis access (ERDA) procedures. We hypothesized that factors independently associated with readmission will evolve from “short-term” to “long-term” determinants at 30-, 180-, and 360-day postoperative cutoff points. Methods: Following IRB approval, medical records of patients who underwent ERDA at two institutions were retrospectively reviewed between 2008 and 2014. Abstracted data included patient demographics, procedural characteristics, the American Society of Anesthesiologists score, Goldman Criteria for perioperative cardiac assessment, the Charlson comorbidity index, morbidity, mortality, and readmission (at 30-, 180-, and 360-days). Univariate analyses were performed for readmissions at each specified time point. Variables reaching statistical significance of P< 0.20 were included in multivariate analyses for factors independently associated with readmission. Results: A total of 450 of 744 patients who underwent ERDA with complete medical records were included. Patients underwent either an extremity revascularization (e.g. bypass or endarterectomy, 406/450) or a noncatheter dialysis access procedure (44/450). Sample characteristics included 262 (58.2%) females, mean age 61.4 ± 12.9 years, 63 (14%) emergent procedures, and median operative time 164 min. Median hospital length of stay (index admission) was 4 days. Cumulative readmission rates at 30-, 180-, and 360-day were 12%, 27%, and 35%, respectively. Corresponding mortality rates were 3%, 7%, and 9%. Key factors independently associated with 30-, 180-, and 360-day readmissions evolved over the study period from comorbidity and morbidity-related issues in the short-term to cardiovascular and graft patency issues in the long-term. Any earlier readmission elevated the risk of subsequent readmission. Conclusions: We noted important patterns in the temporal behavior of hospital readmission risk in patients undergoing ERDA. Although factors independently associated with readmission were not surprising (e.g. comorbidity profile, cardiovascular status, and graft patency), the knowledge of temporal trends described in this study may help determine clinical risk profiles for individual patients and guide readmission reduction strategies. These considerations will be increasingly important in the evolving paradigm of value-based healthcare. Keywords: Morbidity, peripheral vascular interventions, readmission, risk factors, vascular surgery
Although the notion that patient comorbidities and the level of frailty affect surgical outcomes is not novel, the complexities underlying these associations are difficult to study and quantify.[1],[2],[3] This is particularly relevant in vascular surgery, where patients with multiple comorbidities and inherently “high-risk” for readmissions and mortality constitute a norm.[4] As cost containment becomes increasingly important in the evolving health-care delivery paradigm, there is more pressure to reduce postoperative lengths of stay, readmissions, complications, and mortality.[5],[6],[7] Readmissions in vascular surgery patients have been reported in as many as 25% cases,[8],[9] and although not all readmissions are necessarily undesired or preventable,[10] it is certainly an important issue that warrants closer scrutiny and scientific investigation. Identifying “high-risk” patients is a crucial component of the preoperative workup.[11] Current processes tend to be “static” and lack focus on temporal changes in risk profile. A better understanding of how risk factors for readmission evolve during the postoperative period may allow vascular surgeons to more effectively tailor their practice to outcome optimization. Although many risk factors in this domain are known, data of sufficient granularity continue to be limited.[4],[12],[13],[14] Evidence linking comorbidities, degree of arterial disease, length of hospital stay, and the American Society of Anesthesiologists (ASA) physical status to readmissions are compelling but focus mainly on the 30-day postoperative period.[14],[15] It is evident that long-term determinants of readmission are becoming increasingly important, with rehospitalizations beyond the initial postoperative 30-day being both costly and mostly unplanned,[16] yet potentially preventable.[17] Based on the need for more information regarding both short-term and long-term risk of rehospitalization in patients undergoing extremity revascularization or dialysis access (ERDA), we set out to explore in greater detail the relationship between key clinical determinants (demographics, clinical outcomes, and comorbidity indices) and hospital readmissions across a spectrum of temporal milestones. We hypothesized that factors independently associated with hospital readmission will evolve from “short-term” toward “long-term” outcome determinants at 30-, 180-, and 360-days postoperatively.
Study logistics This is a retrospective medical record review of patients who underwent ERDA at two institutions (The Ohio State University Medical Center and St. Luke's University Health Network) between 2008 and 2014. Patients were identified from CPT billing data. Included were patients ages 18–89 who underwent lower extremity revascularization or noncatheter dialysis access (ERDA) procedure. Excluded were patients undergoing aortic and carotid arterial procedures, patients undergoing primary endovascular procedures, repeat/revision bypass or dialysis access operations, central venous catheter placements for dialysis, pregnant patients, prisoners, and those with incomplete medical records. Clinical data were abstracted from hospital records, outpatient clinic electronic records, and follow-up charts as documented by attending physicians, fellows, and residents at both institutions. Study variables Variables collected included patient demographics, procedure details (e.g., type, duration and emergency status), the ASA physical status, Goldman Criteria for postoperative cardiac complications, Charlson comorbidity index (CCI), and congestive heart failure, hypertension, age 75 years, diabetes mellitus, stroke (CHADS2) scores. With the exception of the ASA physical status, all other indices were calculated retrospectively (e.g., patients with incomplete data were not included in the final study sample). Primary data were obtained from the index admission (e.g., the encounter when each patient underwent the initial qualifying vascular procedure). The CCI has been historically used to measure the burden of comorbid diseases.[3],[18] It has been shown to have a consistently strong association with mortality and is an effective means of risk adjustment in outcomes research.[18] Goldman Criteria, ASA, and CHADS2 are other validated clinical tools for perioperative mortality assessment.[12],[19],[20] For the purposes of the current study, postoperative readmission was defined as any inpatient admission at 30-, 180-, and 360-days from the time of discharge following the initial ERDA procedure encounter. Recorded complications during each respective postoperative period were categorized into predefined types (i.e., cardiovascular, cerebrovascular accident/transient ischemic attack [CVA/TIA], graft occlusion or re-occlusion, and any infection). The primary outcome of interest was 30-, 180-, and 360-day hospital readmission following a qualifying ERDA procedure: Either a revascularization (e.g., extremity bypass and endarterectomy) or noncatheter dialysis access (e.g., graft or native arteriovenous fistula creation). Readmissions were analyzed at 30-, 180-, and 360-days against the variables listed above to identify statistically significant univariate associations with primary study outcomes. Mortality was included as a secondary outcome, determined based on a combination of medical record review, and/or examination of the Social Security Death Index (when chart documentation regarding death versus survival was insufficient). Predefined types of complications included the following major categories and subcategories (as applicable): (a) graft occlusion or re-occlusion (excluding stenoses); (b) infection (any infection recorded including wound, urinary, pulmonary, bacteremia/sepsis, and other types); (c) cardiac morbidity (arrhythmia, myocardial infarction, and cardiac arrest); and (d) CVA/TIA (documented CVA or TIA). Variables that reached sufficient statistical significance (P < 0.20) for each respective temporal marker on univariate analysis were subsequently included in the multivariate analysis at the corresponding temporal marker (e.g., 30-, 180-, and 360-days). Statistical analyses The study's ability to detect a 20% difference in primary study outcomes at the significance level α <0.05 and power (1-β) of 0.80 was predicated on a sample size of approximately 400 patients, assuming unmatched group ratios between 2:1 and 4:1. Univariate analyses were performed for readmission as the primary endpoint at each prespecified time marker (30-, 180-, and 360-days). Mortalities during the preceding time period were excluded from subsequent period analyses. Categorical variables were analyzed using the Fisher's exact test. Normally distributed continuous variables were reported as a mean ± standard deviation (unless specified otherwise) and tested using analysis-of-variance. Nonnormally distributed continuous variables were reported as median with interquartile range (IQR) and underwent Mann–Whitney U-testing or Kruskal–Wallis testing as appropriate. Variables achieving statistical significance of P < 0.20 were included in multivariate analyses to determine factors independently associated with readmission for each previously specified time point. Due to relatively small number of events within study outcome categories, as well as the presence of potential “nuisance” parameters, we utilized the backward stepwise logistic regression (likelihood ratio approach) methodology. Statistical analyses were performed using the SPSS 18 software (IBM Corp., Armonk, NY, USA).
Study cohort A total of 450 of 744 patients with complete medical records met inclusion criteria and were entered into this analysis. Patients underwent either an extremity bypass or endarterectomy revascularization (406/450) or noncatheter dialysis access procedure (44/450). Among index procedures, 63/450 (14%) were emergent, and the median operative time was 164 min (IQR 95–217.5 min). There were 188 (41.8%) males and 262 (58.2%) females, with mean age 61.4 ± 12.9 years. Median hospital length of stay (index admission) was 4 days (IQR 2-8 d). Details regarding patient demographic data, comorbidity and physiologic indices, complications, as well as procedure characteristics, can be found in [Table 1].
Factors associated with 30-, 180-, and 360-day readmission Cumulative readmission rates at 30-, 180-, and 360-days were 12%, 27%, and 35%, respectively. Univariate analyses showed that chronologic age and Goldman scores were not significantly associated with readmission at any of the predetermined timeframes. Male gender correlated with readmission at the 30-day mark [Table 2]. The CHADS2 score was associated with readmission at 30- and 180-days but not at 360-days [Table 2]. The CCI, the ASA score, duration of index operation, graft (re)occlusion, and period-specific morbidity (e.g., complications that occurred “between” study temporal markers) correlated with readmission at all predetermined study intervals. Additional results of our univariate analyses are shown in [Table 2].
Multivariate analyses were performed on variables reaching significance level of P < 0.20 on univariate analyses in order to identify factors independently associated with readmission at each temporal study milestone [Table 3]. The CCI and any 30-day morbidity (including in-hospital and postoperative complications) were independently associated with readmission at 30-days. Factors independently associated with readmission at 180-days included the ASA score, graft (re)occlusion, any infection, cardiac morbidity, and 30-day readmission. For the 360-day readmission milestone, graft (re)occlusion, cardiovascular morbidity, and 180-day readmission were among independently associated risks. Details related to the above results, including corresponding odds ratios and 95% confidence intervals are provided in [Table 3]. A diagram showing the temporal evolution of factors associated with readmission can be found in [Figure 1].
The modern health-care delivery paradigm, centered on optimizing quality and value, puts increasing emphasis on the reduction of postoperative hospital stays, readmission rates, complications, and mortality.[5],[6],[7] With readmissions following peripheral vascular surgery procedures affecting as many as 1 in 4 patients, more has to be done to better understand and prevent these occurrences.[8],[9] Our study identified important temporal changes in the behavior of risk factors for readmission in ERDA patients during the 1st postoperative year. More specifically, we examined predictors of readmission at 30-, 180-, and 360-day time points. To our knowledge, similar analyses of long-term readmission patterns have not been published in the past. The current study shows that factors independently associated with readmission at the 30-day postoperative mark include the CCI and 30-day morbidity. These findings suggest that short-term readmissions are most closely linked to the combination of patient comorbidities and postoperative complications. Other authors also emphasize the importance of comorbid conditions in predicting hospital readmissions and outcomes in surgical patients.[21],[22],[23],[24],[25],[26] According to the American College of Surgeons National Surgical Quality Improvement Program data, the following factors were independently associated with 30-day readmission: cardiac comorbidity, chronic obstructive pulmonary disease, dependent functional status, obesity, postoperative stay >6 days, preoperative dialysis, and a preoperative open wound.[22] Jackson, et al.[27] reported that the diagnosis of “critical limb ischemia” and open lower extremity revascularization experienced highest 30-day unplanned rehospitalization rates. In other studies, 30-day readmissions were predicted by a “risk score” combining various “high-risk parameters,” such as demographic characteristics, comorbidities, surgical technical factors, and discharge destination.[14],[28] Our data support the general theme that patient comorbidity burden correlates with readmission risk. Studies traditionally focus on lengths of stay and 30-day readmissions, mainly due to the standardization of these metrics as important surrogates of health-care resource use, care delivery efficiency, and overall quality of the process.[27],[29] However, it has been suggested that the risk of readmission may nearly double following the 90-day postoperative mark, with the chronic disease being one of the most significant risk factors.[30] In the current study, readmissions at the 180-day postoperative marker were independently associated with the ASA score at the time of the original operation, 30-day readmission, graft (re)occlusion, as well as specific (e.g., infection and cardiac) complications. It was interesting to see that the CCI has been supplanted by the ASA score at the 180-day temporal marker [Table 3]. In addition, an earlier readmission predicted additional future readmissions in this study. Literature provides very scant data regarding readmissions beyond the initial 30–90 days postoperatively. In one study, risk factors for readmission within 6 months of the initial infrainguinal bypass operation included the presence of “tissue loss” and “renal failure.”[31] Our data indicate that factors independently associated with readmission at 360-days include vascular graft (re)occlusion, cardiac morbidity, and 180-day readmission. These observations affirm that vascular conduit- and cardiac-related issues, as well as readmission during the preceding study interval (s) dominate the overall risk landscape. In general, the above findings further support the importance of preoperative risk stratification, especially in the context of comorbid conditions, ASA scoring, and cardiovascular status.[13] Despite the paucity of data, other studies of vascular surgery patients seem to support our general findings regarding the risk of readmission at the 1-year temporal marker. More specifically, technical complications requiring additional invasive procedures [32],[33] and cardiac-related events [34] appear to be significantly associated with late readmissions following major vascular operations. Not surprisingly, younger, nondiabetic patients with “near normalization” of postoperative ankle-brachial index have been observed to have better long-term outcomes.[34] We chose to assess comorbidities using established scoring indices previously employed in predicting postoperative mortality and classifying general patient health status.[3],[19],[20],[35] However, none of the scoring systems examined in this study were consistently predictive at all temporal markers [Table 2] and [Table 3]. For example, the CCI was associated with readmissions at the 30-day marker only. Later hospitalizations were characterized by more specific postoperative complications and cardiovascular morbidity, suggesting that not all comorbidities consistently contribute to readmission risk. Of interest, the ASA score was more relevant in determining risk for readmission at 180-days. Although the ASA can be subjective and has some inter-rater inconsistencies, it is certainly more effective in estimating the severity of systemic disease(s) and predicting mortality than simply acknowledging the presence of comorbidities.[36],[37] Regarding longer-term perspectives, other authors have shown that preexisting comorbid conditions may be predictive of adverse events at 6 months postoperatively.[38] Postoperative complications significantly elevate readmission risk. Our results identify any 30-day morbidity, including in-hospital and postdischarge complications, to be independently associated with readmissions. The most significant postoperative complications for readmission at each subsequent stage were graft (re)occlusion and the presence of any infection. These findings imply a need for early prevention, starting with the technical aspects of the procedure, as well as the importance of perioperative care to prevent complications, with the goal of reducing excess readmission rates.[14] A readmission earlier in the postoperative course is also a risk factor for subsequent readmissions and should be incorporated into appropriate risk assessment considerations.[28] As with all outcomes considered to be clinically suboptimal, prevention of hospital readmission should be considered a major health care quality goal, with comprehensive solutions incorporating thorough and appropriate multidisciplinary preoperative evaluation, inpatient care, and outpatient follow-up. Such framework requires that a robust risk assessment system is in place ahead of any scheduled (or even unscheduled, time factors permitting) surgical intervention. Another important factor associated with long-term readmission is perioperative cardiac status. Studies looking at long-term cardiac events and survival in patients undergoing major arterial operations identified “high cardiac risk” to be an independent predictor of late mortality.[38],[39] Although our multivariate analyses point to an association of cardiac morbidity with 180- and 360-day readmissions, the “cardiovascular” specific CHADS2 and Goldman assessment paradigms failed to reach sufficient significance in the current series. The finding that cardiac morbidity is independently associated with readmission risk in this study at both 180- and 360-days, highlights the need for a close, well-coordinated, multispecialty postoperative follow-up. Although we did not specifically address cardiac risk stratification, our results indirectly support the importance of such approach, as exemplified by the Lee index [40] and subsequent management as outlined by Bauer et al.[41] Cumulatively, the evidence suggests that effective identification of long-term risks is critically important to optimizing operative and postoperative course, including readmission risk. This study has several important limitations. First, it is a retrospective endeavor using largely administrative data that heavily depends on the quality of information entered into the combined database. Consequently, some comorbidities, complications, and reasons for readmission may not have been fully characterized, thus contributing to biases in our results. Second, we did not analyze many other potentially influential variables that may have been important factors for readmission (e.g., the Lee index).[40] We did, however, make the best effort to include validated comorbidity-based indices (e.g., CHADS2, CCI, and Goldman).[3],[19],[20],[35] Third, due to limitations of the retrospective medical record review, we were unable to reliably determine for each readmission whether it was planned or unplanned. Thus, we elected to consolidate all planned and unplanned readmissions under the single “readmission” umbrella as a binary outcome. Fourth, there is a component of selection bias inherent to the study, with the exclusion of nearly 300 potentially suitable patients with incomplete medical records. We decided not to use advanced statistical techniques (e.g., multiple imputation) to avoid biases associated with such approaches. Finally, there is probably some patient selection bias originating from the fact that both participating institutions are large teaching hospitals and regional referral centers, with inherently elevated surgical case complexity and acuity. This, in turn, may limit the generalizability of our findings to other clinical settings. Nevertheless, our study is strengthened by the fact that data were provided by two geographically remote institutions (e.g., reducing single-institution bias) and relatively high quality of reporting within included data (e.g., the use of electronic medical records and exclusion of incomplete records) over the duration of each respective patient's initial postoperative year. The ability to accurately assess and understand factors associated with readmissions beyond the “traditional” 30-day period will be increasingly important as various payer sources are likely to expand the “readmission penalty” periods beyond the currently accepted benchmarks.
The current study examined both short-term and long-term risks for readmission in patients who underwent ERDA. While comorbidities play a prominent role in the early readmissions equation, graft-related factors, prior readmissions, and cardiovascular morbidity become increasingly important beyond the initial 30-day postoperative period. Postoperative complications related to the index procedure also appear to modulate the risk of readmission over time. Although our findings are not surprising, they are important and provide a meaningful foundation for readmission prevention. As the value-based health-care paradigm evolves, it is likely that consequences of unplanned rehospitalization will extend well beyond the current, relatively short time horizons. Acknowledgment The authors would like to acknowledge the help of Muntasir Chowdhury and David B. Tulman during the data collection phase of the project. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
[Figure 1]
[Table 1], [Table 2], [Table 3]
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||
