Radical prostatectomy continues to comprise the mainstay of therapy for localized prostate carcinoma. However, caring for radical prostatectomy patients accounts for approximately half of the $1.7 billion annual cost of prostate carcinoma treatment. Length of stay (LOS) after surgery appears to be one of the main components of this cost. The first step in reducing cost is to identify those variables associated with LOS. Radical prostatectomy can be performed using two very different surgical techniques and with each technique different costs are incurred. The objective of the current study was to identify factors associated with LOS as a function of surgical approach. To reduce potential biases due to patient requests for longer hospitalization or physician preferences in that regard, secondary objectives were to identify factors associated with time to fluid intake (TTF) and time to consume solid foods (TTS).
An institutional-based, retrospective chart review of 313 men with clinically localized prostate carcinoma who underwent either a perineal (RPP) or retropubic (RRP) prostatectomy at a single university center from March 1988 to October 1996 was undertaken. Information regarding LOS was available for 311 patients. Linear regression models were used to assess the association between covariables and LOS. Poisson regression models for count data were used to assess associations between covariables and the secondary endpoints of TTF and TTS. Covariables included: preoperative (age, race, prostate specific antigen, Gleason score, clinical stage, lymph node resection, comorbidity, and admission time), intraoperative (surgical approach, surgeon, operative time, estimated blood loss, transfusion requirement, anesthetic approach, and American Society of Anesthesiologists score), and postoperative (pain management complications and transfusions) parameters.
The median LOS was 4 days (range, 1–19 days) for RPP and 5 days (range, 3–16 days) for RRP approaches. The final model included six main effects and three interaction terms. Overall, LOS decreased over time with LOS decreasing at a faster rate in patients who underwent RPP. In general, patients who underwent RRP had an increased LOS compared with patients who underwent RPP. Complications from surgery and age increased the LOS for all patients; however, the increase was greater in patients who underwent RPP. In addition, the use of intraoperative epidural anesthesia and the increased use of postoperative narcotics were associated with increased LOS for patients undergoing both surgical approaches. TTF and TTS were significantly longer for patients who underwent the retropubic approach compared with those patients who underwent the perineal approach. After adjustment for surgical approach no other covariables were found to be associated with TTF. After adjustment for surgical approach, the occurrence of complications was found to be associated with TTS, indicating that patients who experienced complications took longer before they could tolerate solid foods.
Hospital length of stay (LOS) after surgery remains one of the critical components to medical expenditures for a given surgical procedure.1 Radical prostatectomy continues to comprise the mainstay of therapy for patients with localized prostate carcinoma. However, caring for radical prostatectomy patients accounts for approximately half of the $1.7 billion annual cost of prostate carcinoma treatment. LOS after surgery appears to affect this figure substantially. In the present cost-conscious environment of health care, a critical analysis of medical expenditures has led to the alteration of medical practice and the development of care pathways2 that have increased the efficiency of medical practice. Because radical prostatectomy can be accomplished through either a retropubic3 or perineal4, 5 approach, with each being associated with a somewhat different spectrum of side effects,6, 7 variables believed to contribute to LOS need to take the surgical approach used into account.
To address this issue directly, the current study had two objectives: first, to understand the global preoperative and postoperative factors that affect LOS after radical prostatectomy and second, to assess the relevance of these factors as a function of the surgical approach. The study period encompassed the peak incidence rate of prostate carcinoma diagnosis and treatment but ended prior to the initiation of critical care pathways or the significant influx of managed care medicine. Therefore, the current study provided an analysis of factors affecting LOS based solely on physician practice patterns.
MATERIALS AND METHODS
Study Population and Data Collection
Data regarding all men undergoing radical prostatectomy between March 1988 and October 1996 at a single tertiary care university medical center were targeted for analysis. Using information from the institutional billing system, 320 patients with appropriate billing codes for radical retropubic or perineal prostatectomy with or without pelvic lymphadenectomy were identified for chart review. This time interval predates the utilization of a clinical care pathway for total prostatectomy patients established in January 1996 to standardize the care of patients undergoing total prostatectomy at our institution. At the time of data collection, 313 of the 320 patients had medical records allowing for complete review. Due to incomplete documentation of LOS, two additional patients had to be excluded from the LOS analysis. In addition, four other patients were excluded from the time to fluid intake analysis and five patients were excluded from the time to consume solid food analysis because information concerning these data fields was not included in the medical records.
For each patient, information regarding approximately 60 data fields were coded. Of these, 22 variables were prespecified to be of potential predictive value and were classified into preoperative and tumor, intraoperative and postoperative categories. Preoperative and tumor variables included hospital admission time (same day or ≥ 1 day preoperatively and afternoon or morning admission), age, race, prostate specific antigen (PSA) level, prostatectomy Gleason score, clinical and pathologic stage (American Joint Committee on Cancer [AJCC] 1997 version8), lymph node resection, operating surgeon, and physical status classification of the American Society of Anesthesiologists (ASA score)9 as a measure of comorbid conditions. Intraoperative variables included radical prostatectomy approach (perineal or retropubic), surgical time (minutes), estimated blood loss (mL), transfusions (number of units), and anesthetic approach (general or general combined with epidural that was continued for postoperative pain management). Postoperative variables included complications (present or absent), transfusions (number of units), and pain management (intravenous narcotics and/or nonsteroidal10). Transfusions included both autologous and heterologous units.
Endpoints and Statistical Analysis
Endpoints for analysis were LOS, time to fluid intake, and time to consume solid food. Associations between potential predictors and LOS, time to fluid intake, and time to consume solid food were investigated separately. No information regarding patient LOS, time to fluid intake, or time to consume solid food was considered to be a censored observation because partial information was not abstracted (i.e., it was either known or not known at all). Twenty-two prespecified variables were believed to be of potential predictive value. An effect of primary interest was the date of surgery by type of surgical approach interaction because choice of procedure may have changed over time. Linear regression models were used to assess the association between the prespecified covariables and LOS. For modeling purposes the natural logarithmic variance stabilizing transformation of LOS was used. Two hundred twenty-two of the patients (71%) had complete documentation in all data fields.
Several methods of variable selection, including backward and stepwise (based on reduction in Mallow's Cp statistic)11 in conjunction with the bootstrap,12 were used to determine which main effects were associated with LOS. All relevant main effects by surgical approach interactions and prespecified biologically feasible interactions were investigated. The subset of patients with complete data was used for initial model building with variables being retained in the model with a two-sided P value < 0.1. Patients with complete information regarding all retained variables were added to the final descriptive model and it is these P values that are reported in the current study. No noticeable changes in the coefficient values were observed. Regression diagnostics11 were applied to detect outliers and to assess the relative influence of each patient on the model. Variables whose significance was dependent on a few highly influential cases were deleted from the final model. In addition, nonparametric Cox proportional hazards models13 with stepwise variable selection were used to model LOS as a function of the covariables. Identical covariates were found to be associated with LOS. Poisson regression models14 for count data were used to assess associations between time to fluid intake and time to consume solid food as a function of the prespecified covariables. Selection methods similar to those used for LOS were employed.
Patient demographics, tumor characteristics, and preoperative information by surgical approach are summarized in Table 1. Patient characteristics were similar between the two surgical approaches. A majority of the patients were white (88%) and the overall median age was 65 years (range, 43–83 years). The majority of patients had pathologic T2 disease (71%). The median preoperative PSA was 8.1 (range, 0.3–98.0) and the median Gleason score was 6 (range, 2–9). Depending on which surgical approach was used, differing percentages of patients were observed to have undergone lymph node resection; 183 patients undergoing a radical retropubic prostatectomy (99%) underwent lymph node resection whereas only 7 patients undergoing a radical perineal prostatectomy (6%) underwent lymph node resection. The distribution of operating surgeon differed by surgical approach. Within each approach a different surgeon performed the majority of surgeries.
Intraoperative and postoperative information is summarized in Table 2. In general the perineal approach required less surgical time (a median of 184 minutes) than did the retropubic approach (a median of 259 minutes). In addition, estimated blood loss (median of 500 mL vs. 1200 mL, respectively) and epidural use (25% vs. 59%, respectively) were less with the perineal approach compared with the retropubic approach. Complication rates were similar between the two approaches; however, narcotic use for pain management tended to be higher in those patients who underwent the retropubic approach. Transfusions (either autologous or heterologous) were given to a greater percentage of patients in the retropubic group compared with the perineal group (84% vs. 11%, respectively).
Table 2. Intraoperative and Postoperative Variables
The median LOS was 4 days (range, 1–19 days) for the perineal approach and 5 days (range, 3–16 days) for the retropubic approach. The LOS for perineal prostatectomy was significantly less (P < 0.05) than for radical retropubic prostatectomy (Fig. 1). Without adjustment for surgical approach, increased surgical time, epidural use, higher estimated blood loss, more postoperative narcotic use, surgical complications, and more transfusions were singularly associated with an increased LOS (P < 0.05).
An effect of primary interest was the LOS as a function of the type of surgical approach and date of surgery because patient discharge trends may have changed over time. This effect was modeled with a restricted cubic spline (Fig. 2). Overall, LOS decreased over time (P = 0.011) with LOS decreasing at a faster rate in those patients who underwent the perineal approach. Linear regression models were used to assess the association between covariables and transformed (i.e., ln[LOS]) LOS. The final model included six main effects and three interaction terms. The six main effects were surgical approach (P < 0.001), year of surgery (P < 0.001), epidural use (P = 0.005), surgical complications (P < 0.001), age (P = 0.027), and postoperative narcotic use (P < 0.001). Interactions between type of surgical approach and year of surgery (P = 0.004), age (P = 0.071), and complications (P = 0.001) were found to be associated significantly with LOS.
In general, patients who underwent the retropubic approach had an increased LOS compared with patients who underwent the perineal approach. Complications from surgery and age increased the LOS for all patients; however, the increase was greater in patients who underwent the perineal approach. The use of intraoperative epidural anesthesia and the increased use of postoperative narcotics were associated with an increased LOS in patients treated by either surgical approach. The relative importance of these predictors is shown in Figure 3.
The median time to fluid intake in days was 0 (range, 0–4 days) for patients undergoing the perineal approach and 1 day (range, 0–9 days) for patients undergoing the retropubic approach. Fluid intake occurred sooner (P < 0.05) in patients who underwent the perineal approach (Fig. 4). The median time to solid food consumption in days was 1 day (range, 0–8 days) for those patients who underwent the perineal approach and 3 days (range, 0–11 days) for those patients who underwent the retropubic approach. Similarly, the ability to consume solid foods occurred sooner (P < 0.05) in patients who underwent the perineal approach (Fig. 5). As seen in Figures 4 and 5, there was little dispersion in time to fluid intake and time to solid food consumption in these data. Therefore, our ability to distinguish associations between covariables was limited. Time to fluid intake and time to consume solid foods were significantly longer for patients who underwent the retropubic approach compared with those who underwent the perineal approach. After adjustment for surgical approach, no other covariables were found to be associated with time to fluid intake. However, after adjustment for surgical approach, the risk of whether a patient experienced complications was found to be associated with time to consume solid foods. Patients who experienced complications took longer before they were advanced to a solid diet. It is interesting to note that, unlike LOS, year of surgery was not found to be associated significantly with either time to fluid intake or solid food consumption, possibly indicating that the primary reason for the decreasing LOS over time was the more rapid discharge of patients who already were receiving a regular diet.
In view of the significant yearly incidence rate of new cases of localized prostate carcinoma,15 the costs associated with potentially curative treatment of the disease at this stage comprise a significant portion of health care expenditures. Currently, the potentially curative treatment modalities offered to patients with clinically organ-confined prostate carcinoma include radical prostatectomy,3 external beam radiation therapy,16 and interstitial brachytherapy.17 Despite the increased use of brachytherapy, radical prostatectomy continues to be the most common18 and perhaps the most definitive19, 20 long term curative therapy for patients with this stage of disease.
A radical prostatectomy can be accomplished through either a retropubic3 or perineal4, 5 approach. Based on the number of recent literature citations, the perineal approach to radical prostatectomy has regained some of the popularity6, 21–23 it enjoyed in the past. Data from several groups indicate excellent freedom from PSA recurrence in patients treated with this approach.6, 21 Although Harris and Thompson21 reported that PSA is detectable postoperatively in 2% and 4% of pT2 and pT2 to T3b cases, respectively, at an average follow-up of 1 year, a recent study demonstrated that care must be taken in the complete excision of the seminal vesicles for biochemical failure rates to remain comparable to those of the retropubic approach.6 A disadvantage associated with this technique is the increased rate of incidence of fecal incontinence.7 Conversely, this technique offer some advantages with respect to urinary incontinence over that accomplished with the retropubic approach,7 but this may be applicable only to minor degrees of incontinence that do not impact the patient's urinary quality of life.24
In 1985 the reported LOS for patients undergoing total prostatectomy was > 2 weeks, with a mean of 15 days for the perineal approach and 19 days for the retropubic approach.25 By 1993, with a better understanding of the surgery and improvements in intraoperative technique and anesthetic care, several groups reported a LOS of on average of 5 days.1, 26, 27 This was decreased further by the establishment of clinical care pathways2 for these procedures, which has resulted in a LOS that approaches a 24-hour stay.28 To our knowledge this decrease in LOS has not been associated with increased complications1 or patient dissatisfaction.29
Both the positive findings and the negative findings of the current study provide insight into the factors that may and may not effect LOS after radical prostatectomy. This analysis of LOS in a patient population in which two different approaches to the same surgery were used affords an excellent opportunity to evaluate the factors that are important for each approach and thus may allow a certain degree of tailoring of the surgical approach to the preoperative status of the patients. Because the tumor characteristics of each group were similar at the time of the procedure, the differences in LOS after adjustment for the surgical approach allows a window into many of the possible operative or postoperative events that may affect LOS after total prostatectomy. These results also are of value in settings in which only one form of the procedure is performed because they define the risk factors associated with the LOS for each procedure and thus help in the preoperative counseling of patients.
In an era free of managed care or clinical care pathways, the steady decline in LOS while the time to fluid intake and time to consume solid food remained constant that was observed in the current study for both approaches suggests that physician and patient preferences for shorter hospital stays may underlie some of these findings. The results of the current study demonstrate that the variables that impact LOS vary according to surgical approach. Complications from surgery and the age of the patients increased the LOS for all patients; however, the increase was greater in patients undergoing the perineal approach. In addition, it appears that epidural anesthesia when combined with general anesthesia has an unfavorable impact on LOS while not affecting time to fluid intake and time to consume solid food. One possibility that may account for this finding is that the epidural anesthesia was affecting the ambulatory status of the patients, which slowed hospital discharge although it did not affect bowel function. These data are consistent with those of Stevens et al.,30 who showed that combined epidural anesthesia and “light” general anesthesia with spontaneous ventilation decreased intraoperative blood loss and shortened the time to the return of bowel function. However, in the current study the earlier return of bowel function was not great enough to realize a difference in LOS. This potential negative effect of epidural use on LOS must be placed in perspective with the fact that the use of preemptive epidural analgesia is associated with significantly decreased postoperative pain during hospitalization and long after discharge and with increased activity levels after discharge.31
By performing a comprehensive analysis of factors contributing to hospital LOS after radical prostatectomy, the current study demonstrated that the impact of such factors is highly dependent on the surgical approach used. In view of the importance of clinical care pathways in reducing medical expenditures from radical prostatectomy, the results of the current study may contribute to the further refining of these pathways by highlighting the differences and similarities among the variables affecting LOS as a function of surgical approach.