Cost-utility analysis of negative pressure wound therapy in high-risk cesarean section wounds

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

1 2 3 4 5 6 7 8 9 10 11 12 13 Q1 14 15 16 Q12 17 Q2 18 19 Q3 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.JournalofSurgicalResearch.com

Cost-utility analysis of negative pressure wound therapy in high-risk cesarean section wounds Haitham W. Tuffaha,a,b,c,* Brigid M. Gillespie, PhD,a,c Wendy Chaboyer,a,c Louisa G. Gordon, PhD,a,b and Paul A. Scuffhama,b a

Griffith Health Institute, Griffith University, Gold Coast, Queensland, Australia Centre for Applied Health Economics, School of Medicine, Griffith University, Meadowbrook, Queensland, Australia c NHMRC Centre of Research Excellence in Nursing Interventions for Hospitalised Patients, Research Centre for Health Practice Innovation, Griffith University, Gold Coast, Queensland, Australia b

article info

abstract

Article history:

Background: Obese women undergoing cesarean section are at increased risk of post-

Received 29 October 2014

operative infection. There is growing interest in negative pressure wound therapy (NPWT)

Received in revised form

to prevent closed surgical incision complications including surgical site infection; however,

14 January 2015

the evidence on the effectiveness and cost-effectiveness of this technology is limited. The

Accepted 6 February 2015

objective of this study was to evaluate the cost-effectiveness of NPWT compared with that

Available online xxx

of standard dressing in preventing surgical site infection in obese women undergoing elective cesarean section based on current evidence and to estimate the value and optimal

Keywords:

design of additional research to study this technology.

Cost-effectiveness

Methods: The analysis was from the perspective of Queensland Health, Australia, using a

Negative pressure wound therapy

decision model. Parameters were obtained from the published literature, a pilot clinical

Cesarean section

trial, and expert opinion. Monte Carlo simulation was performed to calculate the net

Value of information

monetary benefit, characterize decision uncertainty, and estimate the value of additional research. Comparing the expected monetary benefits and costs of alternative trial sample sizes informed the optimal future study design. Results: The incremental net monetary benefit of NPWT was Australian dollars 70, indicating that NPWT is cost-effective compared with that of standard dressing. The probability of NPWT being cost-effective was 65%. The estimated value of additional research to resolve decision uncertainty would be Australian dollars 2.7 million. The optimal sample size of a future trial investigating the relative effectiveness of NPWT would be 200 patients per arm. Conclusions: Based on the current evidence, NPWT is cost-effective; however, there is high uncertainty surrounding the decision to adopt this technology. Additional research is worthwhile before implementation. ª 2015 Elsevier Inc. All rights reserved.

1.

Introduction

The increasing prevalence of obesity in women of childbearing age is a major health problem. Studies from the United States,

England, and Australia reported around 25% of women of childbearing age are obese with a body mass index (BMI) of
30 kg/m2 [1e4]. Maternal obesity poses serious complications during and after pregnancy to both the affected mothers and

* Corresponding author. Centre for Applied Health Economics, School of Medicine, Griffith Health Institute, Griffith University, Queensland 4131, Australia. Tel.: þ61 7 338 21510; fax: þ61 7 338 21338. E-mail address: [email protected] (H.W. Tuffaha). 0022-4804/$ e see front matter ª 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jss.2015.02.008

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130

2

131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 Q8 182 183 184 185 186 187 188 189 190 191 192 193 194 195

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

their babies, including gestational diabetes, hypertensive disorders of pregnancy, and stillbirth [5]. Obesity also increases the need for cesarean delivery with the risk of a cesarean section (CS) being two to three times higher among obese compared with pregnant women of normal weight [6e8]. Obese women undergoing CS are at increased risk of complications particularly postoperative infection [5,9]. In a metaanalysis of six studies, the pooled odds ratio for obese CS women having an infection was three times higher compared with that for nonoverweight women [5]. A common postoperative complication is surgical site infection (SSI), which occurs after surgery in the area of the body where the surgery took place [10]. Controlling SSI is a health-care quality indicator because it results in significant morbidity, reduced quality of life, occasional death, and increased costs [11e13]. One case of SSI may cost up to $30,000, depending on its severity [12,14,15]. Despite the advances in infection control practices, ventilation systems in the operating rooms, sterilization methods, surgical technique, preoperative antimicrobial prophylaxis, and wound dressings, SSI remains common in obese women undergoing CS with an estimated incidence between 16 and 30% [11,16,17]. Since its introduction two decades ago, negative pressure wound therapy (NPWT) has been used to promote the healing of acute and chronic wounds as well as skin grafts (Table 1) [19e21]. It is based on a closed sealed system that applies negative pressure to the wound surface resulting in increased blood circulation, decreased edema, enhanced granulation tissue formation, and reduced bacterial colonization [21,22]. There is growing interest in extending the use of NPWT to closed surgical incision to prevent wound complications including SSI [21,22]. Unfortunately, the available evidence on the effectiveness and cost-effectiveness of NPWT in surgical incisions is limited [23]. This is expected in surgical practice, where innovations in technologies and equipment often outpace supporting evidence. Recent systematic reviews have identified three small randomized controlled trials (RCTs) that investigated the incidence of SSI in NPWT compared with that of standard wound dressing [22,23]. Those trials showed a reduction in SSI with NPWT although all trials reported that the reductions were not statistically significant [23e26]. None of the trials involved patients undergoing CS.

Given the cost of NPWT can reach $100 a day, it is essential to evaluate the cost-effectiveness of this technology before its wide implementation. Nevertheless, with the limitations in the available evidence, the results of a cost-effectiveness analysis may not be certain enough to inform a decision. Clearly, conducting additional research would reduce this uncertainty and better inform decisions. But, there is a cost associated with obtaining further evidence in terms of the direct costs of conducting clinical trials and the opportunity cost of delaying the implementation of an effective intervention awaiting research results. An analytical approach known as value of information analysis has been developed and used in health-care interventions to inform whether the available evidence is sufficient to support a decision on a given technology or if additional research study is worthwhile [27,28]. It is based on the notion that information is valuable because it reduces the uncertainty surrounding the available evidence and subsequently the potential cost of making wrong decisions based on that uncertain evidence [27,28]. In other words, the expected value of information is the expected cost of error. Furthermore, value of information analysis has been proposed as an alternative to the standard hypothesis testing approach, which is based on type 1 and type 2 error and the minimum clinically important difference, in determining sample sizes for RCTs [29e31]. Under this economic approach, researchers consider the sample sizes that maximize the expected net benefit of research, which is the difference between the expected monetary benefit of a given trial design and its expected cost [29,30]. The aim of this study was to conduct a cost-effectiveness analysis of NPWT in preventing SSI in obese women undergoing CS compared with that of standard dressing based on currently available evidence and to perform a value of information analysis to estimate the value and optimal sample size of a larger RCT to support this technology.

2.

Methods

The approach to achieve the study aim was to: 1) conduct a cost-effectiveness analysis of NPWT compared with standard dressing using a decision analytic model and; 2) perform

Table 1 e Description of commonly used negative pressure devices [18]. Productname Manufacturer Clinical indications

Pressure settings Therapy duration, d Cost

Prevena Kinetic Concepts Inc  Chronic wounds  Acute wounds  Traumatic wounds  Subacute wounds  Dehisced wounds  Partialthickness burns  Flaps and grafts 75 to 125 mm Hg 7 AUD875

VAC-VIA Kinetic Concepts Inc Clean, closed incisions that continue to drain after closure.

125 mm Hg 2e7 AUD395

mm HG ¼ millimeter mercury.

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

PICO Smith and Nephew  Acute  Flaps and grafts  Incision sites  Partial thickness burns  Subacute wounds  Traumatic  Ulcers (e.g.,pressure) 80 mm Hg 5 AUD175

196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325

Monte Carlo simulation to characterize decision uncertainty and estimate the expected value of additional research.

2.1.

Cost-effectiveness analysis

The cost-effectiveness analysis was from the perspective of the State Department of Health in Queensland, Australia, using a decision model. The model was probabilistic with prior distributions assigned to input parameters. We used Monte Carlo simulation to sample from the input distributions to estimate the expected costs and effects associated with each intervention [32]. In general, beta distributions were assigned to probabilities and utilities, gamma distributions to costs and disutilities, and lognormal distributions to relative risks (RRs). For this analysis, the efficacy outcome was qualityadjusted life-years (QALYs) gained. The net monetary benefit (NB) was calculated, which is the efficacy multiplied by the willingness-to-pay threshold for additional unit of effect outcome minus the cost [33]. We set the willingness-to-pay threshold at Australian dollars (AUD) 50,000 per QALY [34]. The intervention expected to be cost-effective would be the one with the highest expected monetary net benefit. Costs and net benefits were presented in AUD. The time horizon of the model was 6 mo to allow for sufficient time to capture and treat post-CS complications. Costs and effects were not discounted because the model timeline was <1 y. The output of the simulation was used to characterize decision uncertainty presented as the probability that each treatment has the highest expected monetary net benefit.

2.1.1.

The decision model

To describe the clinical problem, we constructed a decision tree (Fig. 1) in TreeAge Pro 2013 (TreeAge Software Inc,

3

Williamstown, MA) to show the outcomes of a hypothetical group of obese women (i.e., BMI
30 kg/m2 before pregnancy) with an average age of 32 y who underwent elective CS. The two groups would receive the same antibiotic prophylaxis before surgery and would be operated using the same technique and under the same setting. At the completion of skin closure, NPWT would be applied to one group and the other group would receive standard dressing (i.e., hydrocolloid). The modeled patients may develop SSI, which could be either superficial or deep and/or organ. To simplify the model, deep and organ SSI were combined as deep/organ SSI. Superficial SSI occurs within 30 d after the operation and only involves skin and subcutaneous tissue of the incision [10]. Deep SSI occurs within 30 or 90 d after the operation and involves deep soft tissues of the incision (e.g., fascial and muscle layers) [10], whereas organ SSI involves any part of the anatomy (e.g., organs and spaces) other than the incision, which was opened or manipulated during an operation [10]. Patients could die or survive depending on the type of the SSI developed. Death from other causes (e.g., age-related death or death from surgery) was not included because the probability of death in this young group of patients undergoing such procedure is minimal [35].

2.1.2.

Parameters for use in the model

Parameters were obtained from a systematic review of literature (Table 2). Expert opinion was sought when a parameter value could not be found in the published articles. Furthermore, to ensure that all relevant evidence was included in the model, data from a recent pilot study conducted by our group were included and combined with already available evidence. The details of the pilot trial are published elsewhere [36]. In brief, that trial assessed the effect of NPWT on SSI in obese women undergoing elective CS in addition to the feasibility of

Fig. 1 e Economic model structure. (Color version of figure is available online.) 5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390

4

391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

Table 2 e Model input parameters. Parameter Probabilities, % SSI with standard dressing RR of SSI with NPWT Deep/organ SSI Death from superficial SSI Death from deep/organ SSI Utilities Utility with no SSI Disutility from superficial SSI Disutility from deep SSI Utility of death Costs, $ NPWT Standard dressing Superficial SSI treatment Deep/organ SSI treatment Application time (h) NPWT Application time (h) dressing

Value

Distribution

Source

24

Beta (a ¼ 125, b ¼ 381)

72 20 4 9

Log-normal (0.33 to 0.33) Beta (a ¼ 90, b ¼ 356) Beta (a ¼ 37, b ¼ 845) Beta (a ¼ 41, b ¼ 410)

Johnson et al. [11], Opoien et al. [35], Ahmed et al. [16], and Alanis et al. [17]. Pilot trial [36] and Masden et al. [25]. Wilson et al. [13] and Henman et al. [37]. Astagneau et al. [38] and Kirkland et al. [39]. Astagneau et al. [39] and Kirkland et al. [38].

Beta (a ¼ 185, b ¼ 18) Gamma (a ¼ 8, b ¼ 41) Gamma (a ¼ 16, b ¼ 40) Fixed

Clemens et al. [40]. Lipsky et al. [41]. Lipsky et al. [41]. Assumed

Fixed Fixed Gamma Gamma Gamma Gamma

Market price Market price Graves et al. [42]. AR-DRGs [43]. Expert estimate Expert estimate

0.91 0.2 0.4 0 175 7.5 250 10,000 0.15 0.05

(a (a (a (a

¼ ¼ ¼ ¼

4, b ¼ 0.02) 4, b ¼ 0.0004) 9, b ¼ 60) 40, b ¼ 40)

AR-DRGs ¼ Australian Refined Diagnosis Related Groups.

conducting a definitive trial. Ninety-two obese women undergoing elective CS from July 2012eApril 2014 were randomized in theater (i.e., at the completion of skin closure) via a central Web-based system using a parallel 1:1 process to Q4 Q5 either NPWT (PICO) or standard dressing (Comfeel Plus). In 420 total, 27.9% of the control group and 22.7% of the intervention 421 group had an SSI with a RR of 0.81 (95% confidence interval [CI] 422 0.39e1.68) [36]. Based on the pilot study, a larger RCT to test 423 the superiority of NPWT in reducing SSI incidence appeared 424 feasible, with an estimated sample size of 400 patients per 425 arm with over 90% power [36]. 426 427 2.1.2.1. Probabilities. Baseline risk of SSI in the group 428 receiving standard dressing was set at 24%, estimated from 429 the incidence of SSI control arm in the pilot trial combined 430 431 with four observational studies reporting SSI in obese women 432 undergoing CS [11,16,17,35]. 433 No RCTs comparing NPWT with standard dressing in CS 434 were identified. One RCT by Masden et al. [25] reported the 435 relative effectiveness of NPWT in reducing SSI in 81 high-risk 436 patients (i.e., BMI >30 and comorbidities) undergoing a range 437 of procedures including abdominal surgeries; 6.8% of the 438 NPWT group and 13.5% of the standard dressing group 439 developed wound infection with an RR of 0.50 (95% CI 440 0.13e1.95). Other trials identified were an RCT by Howell et al. 441 [24] on NPWT in knee surgery that was terminated early due to 442 443 blister formation, and an RCT by Stannard et al. [26] investi444 gating NPWT in 249 patients with lower extremity trauma 445 fractures. In that RCT, around 10% of wounds in the NPWT 446 group had infection compared with 20% in the standard 447 dressing group at an RR of 0.52 (95% CI 0.28e0.96) [26]. It was 448 not appropriate to combine the results from Stannard et al. 449 with Masden et al. because the analysis in the former was per 450 wound and not per patient. Furthermore, patient character451 istics and wound types in the two studies were heteroge452 neous. Given the scarcity in the available evidence and in 453 454 order not to overestimate uncertainty in the relative

effectiveness parameter by relying on the pilot study results alone, the RR from the pilot study was collated with the RR from Masden et al. This was achieved by undertaking a Bayesian approach under which the RR from Masden et al. (i.e., prior information) was updated with the RR from the pilot trial resulting in an updated (i.e., posterior) RR of 0.73 (0.39e1.32) [44]. The effect of RR estimation on the results of the costeffectiveness and value of information analyses was explored in sensitivity analysis. The probability for deep/organ SSI was estimated at 19% from Wilson et al. [13] and Henman et al. [37]. The probability of death from deep/organ SSI was set at 0.07 and for superficial SSI at 0.02, from Astagneau et al. [39] and Kirkland et al. [38].

2.1.2.2. Costs. The cost of NPWT was set at AUD175 for the price of a disposable (one-application) device (PICO). The cost of standard dressing was AUD7.5 for the hydrocolloid dressing (Comfeel Plus). The cost of treating superficial SSI was obtained from Graves et al. [42] and was set at AUD250; this includes the cost of a general practitioner visit, 7 d of oral antibiotic, and the cost of test and/or swab. For the cost of deep/organ SSI, this was obtained from the 2009e2010 Australian Refined Diagnosis Related Groups, item T61 (postoperative and posttrauma infection) at AUD10,000 [43]. This includes the cost of hospitalization, tests and/or swabs, and intravenous antibiotics for 7e14 d [43]. The estimated staff time was 10 min to apply the NPWT and 2 min for the standard dressing at an average wage of AUD33 per hour [45]. Costs obtained in other price years were converted to 2014 AUD using the CCEMG-EPPI-Centre Cost Converter Web-based tool [46]. 2.1.2.3. Utilities. The utilities in the model were based on EuroQoL 5D (EQ-5D-3L) scores, anchored between 0.0 for death and 1.0 for best possible health. Utility weights were based on the preferences of the Australian population. The utility scores for the women undergoing CS and discharged with no complications was set at 0.9 from Clemens et al. [40]. For the

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519

5

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584

women who developed SSI, the disutility for superficial and deep/organ SSI was set at 0.2 and 0.4, respectively, from Lipsky et al. [41]. The assumption was that the disutility will take place over 1 wk for superficial SSI and 2 wk for deep/organ SSI.

2.2.

Value of information analysis

The detailed algorithms for value of information calculation are described in the literature [47,48] and are presented in Appendix 1. Briefly, the first measure to calculate was the expected value of perfect information (EVPI). This is the value of the additional information that would resolve all uncertainty surrounding “all” input parameters; therefore, it is the maximum (upper bound) value for conducting further research to resolve this uncertainty [44]. The EVPI is the difference between the expected NB of a decision with perfect information and the decision made based on current information [49]. To estimate the EVPI, 10,000 Monte Carlo iterations were randomly sampled from the prior parameter distributions of all parameters to identify the intervention with the highest expected mean NB based on current information (i.e., the cost-effective intervention). Then the intervention with the highest NB “at each iteration” was identified and the identified values were averaged to calculate the expected “maximum” NB (i.e., NB with perfect information). If the EVPI exceeds the expected cost of future research, the next step would be to calculate the value of information to resolve the uncertainty in the parameter(s) of interest, which is the expected value of perfect parameter information (EVPPI) [50,51]. Since our model (i.e., the decision tree) is linear and assuming no correlation between input parameters because they were obtained from various sources, the same one-level Monte Carlo simulation technique described previously was used to calculate the EVPPI; the sampling would be only from the distribution of the “parameter(s) of interest,” whereas the other parameters were fixed at their prior means [51]. To estimate the value of a future clinical trial with a given sample size (n) that could reduce uncertainty surrounding the parameter of interest, the expected value of sample information (EVSI) was estimated by calculating the difference between the expected value of a decision made after collecting data on the parameter of interest and the expected net benefit with current information [47]. Conceivably, the data collected from additional research are not known at this stage but could be predicted by simulation. Given the linearity of the model, calculating the EVSI required the same one-level Monte Carlo simulation. However, the sampling would be from the posterior distribution of the parameter(s) of interest obtained using Bayesian updating [47]. The value of information measures described previously are per-patient estimates; however, it is necessary to estimate the value of information for the population of patients expected to benefit from the research outcomes. This was calculated by multiplying the per-patient estimates by the estimated number of patients expected to benefit from NPWT over a certain period. Obese women undergoing CS in Queensland represent 20% of the 20,000 CS performed every year in that state [43,52]. Accordingly, we estimated the expected number of obese women undergoing CS over 10 y (with 5% discounting) to be around 35,000.

To determine the optimal sample size of a future trial, the population-EVSI and the expected total cost were estimated for a range of possible trial sample sizes. The difference between the expected monetary benefit of research and the total cost of a particular study design is the expected net benefit of sampling (ENBS) [31,53]. The total cost of a future trial design included fixed costs (e.g., set-up cost, salaries), variable costs per patient, and the opportunity costs expected to be incurred by patients who would receive the inferior intervention during the trial [47,53]. We based our estimates for the cost of a future trial design on a research grant application for an RCT on NPWT in four institutions with a recruitment rate of 200 patients per site each year (Table 3). The estimated fixed project cost was AUD125,000 per year for project management and data analysis, plus an annual cost of AUD 100,000 per site for recruitment and data collection. The cost per patient was set at AUD250. If the ENBS is negative, additional research would not be cost-effective because the expected costs of the study would exceed its expected benefits. Conversely, a positive ENBS indicates that future research would be worthwhile. The optimal sample size is determined when the ENBS reaches a maximum [54,55].

3.

Results

3.1.

Cost-effectiveness analysis

Compared with standard dressing, NPWT resulted in an average additional cost of AUD30 (AUD600 versus AUD570) and additional 0.002 QALYs (Table 4). At a willingness-to-pay threshold of AUD50,000 per QALY, the incremental NB was 70AUD, indicating that NPWT is cost-effective. The probability of NPWT being cost-effective was 65%. Figure 3 shows the probability of NPWT being cost-effective over a range of willingness-to-pay thresholds.

3.2.

Value of information analysis

The EVPI for the decision of adopting NPWT is AUD76 per patient, which is AUD2.7 million (AUD76  35,000) for the population expected to benefit from this technology over the coming 10 y. The parameter with the highest value of information was the RR of SSI with NPWT at AUD75 per patient and population value of AUD2.6 million. The value of a future RCT exploring the relative effectiveness of NPWT over a range of

Table 3 e Research cost breakdown. Item Fixed costs Data management/y Project management/y Office supplies/y Field expenses (i.e., site visits and monitoring)/site/y Recruitment and data collection salaries/site/y Blinded outcome assessor/site/y Variable cost (i.e., per patient) Equipment Randomization services

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

Cost (AUD) 30,000 85,000 10,000 15,000 70,000 15,000 200 50

585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649

6

650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

Table 4 e Cost-effectiveness analysis results. Analysis results

Standard dressing

NPWT

Difference

Cost (AUD) Effect (QALY) NB* (AUD)

570 0.446 21,730

600 0.448 21,800

30 0.002 70

*

For a willingness-to-pay threshold of AUD50,000 per QALY.

sample sizes per arm is depicted in Figure 3. As the sample size increases more uncertainty is expected to resolve and the value of additional research increases. Comparing the expected monetary benefits and costs of the suggested sample sizes, the optimal sample size would be 200 patients in each arm with an ENBS of AUD1.2 million at a total cost of AUD900,000 (Table 5). The expected return on investment (i.e., net benefit/cost ratio) would be 133% (AUD1.2million/ AUD900,000). The initial design with 400 patients per arm would provide a return on investment of 66% (AUD935/ AUD1.4million). In a sensitivity analysis, increasing the price of NPWT, varying willingness-to-pay threshold, extending the timeline of the technology, or estimating the RR based on the pilot trial alone resulted in an estimated optimal sample size between 200 and 300 patients in each arm (Table 6). On the other hand, with reduced NPWT price and shorter technology life-time the estimated sample sizes ranged between100 and 200 patients in each arm.

4.

Discussion

The use of NPWT in closed surgical wounds to enhance healing by primary intention and to prevent wound

complications is a new field of application for this technology. This article presents a cost-effectiveness analysis of NPWT in preventing SSI in obese women undergoing CS. Based on the current evidence, NPWT appears to be cost-effective compared with standard dressing with an expected incremental NB of AUD70. Nevertheless, the probability of NPWT being cost-effective is only 65%, indicating high decision uncertainty and thus high chance of error in a decision based on this cost-effectiveness analysis. Given the high cost of NPWT and the high uncertainty in the cost-effectiveness results, it would be reasonable to conduct additional research before implementing this technology. The expected value of information to resolve the uncertainty in the available evidence would be around AUD2.7 million, suggesting that additional research is potentially worthwhile. Our analysis estimates the optimal sample size for a future trial investigating the relative effectiveness of NPWT compared with standard dressing in reducing SSI. By calculating the expected monetary benefit (i.e., the expected reduction in uncertainty) of additional sampling and the expected cost of conducting this future trial, the sample size with the highest benefit-to-cost ratio would be 200 patients in each arm. This sample size is lower than the sample size of 400 patients in each arm initially calculated based on hypothesis testing and this smaller sample size would be more economical providing higher return on investment (133% versus 66%). The results demonstrate how value of information analysis can provide an alternative to the standard hypothesis testing approach, which relies on arbitrary chosen error probabilities where type 1 and type 2 error receive the same weight (e.g., 5% and 20%, respectively), regardless of the consequences of making an error [30]. Under value of information analysis, an economic approach is applied to sample size estimation. This approach considers a number of factors

Fig. 2 e The probability of each intervention being cost-effective over a range of willingness-to-pay thresholds. (Color version of figure is available online.) 5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

Q9

715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779

7

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844

Fig. 3 e The expected value of information and expected cost across future trial sample sizes. (Color version of figure is available online.)

such as the relative effectiveness and costs of the evaluated technologies, the decision maker’s willingness-to-pay for the additional effectiveness, the probability and consequences of making a suboptimal decision, the population expected to benefit from research findings, the level of implementation (i.e., uptake) of research findings, and the total cost associated with the intended research [54]. Of note, the total cost of research does not only include the direct cost of research in terms of fixed and variable costs but also the opportunity cost from delaying the implementation of the technology awaiting the conclusion of the future trial [27,56]. In addition to sample-size calculation, value of information analysis can optimize other aspects of trial design such as the number of comparators and follow-up duration [29,56]. Obviously, more uncertainty is expected to resolve with longer

follow-up and more comparator arms albeit with additional research costs. Accordingly, the preferred design would be the one that optimizes the expected research monetary benefits compared with the expected research costs [53]. The same principle can be extended to quantitatively prioritize research. Under the value of information framework, competing research proposals within a limited budget could be ranked according to their expected return on investment [53]. To our knowledge, there is no published cost-effectiveness analysis of NPWT in preventing wound complications in closed surgical incisions [23]. There is, however, a limited number of published studies evaluating the cost-effectiveness of NPWT in the management of chronic and open wounds [55,57e59]. The lack of robust clinical evidence (i.e., large RCTs) to support NPWT may explain the rarity of relevant

Table 5 e Expected cost, benefits, and return on investment for future trial design. Sample size/arm 100 200 300 400 500 600 700 800 900 1000

EVSI (AUD)

Research sites number

Trial duration (y)*

Total trial costy (AUD)

ENBSz (AUD)

ROI, %x

1,645,000 2,114,000 2,275,000 2,345,000 2,380,000 2,415,000 2,432,500 2,448,250 2,457,350 2,460,500

4 4 4 4 4 4 4 4 4 4

1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50

656,250 907,500 1,158,750 1,410,000 1,661,250 1,912,500 2,163,750 2,415,000 2,666,250 2,917,500

988,750 1,206,500 1,116,250 935,000 718,750 502,500 268,750 33,250 208,900 457,000

151 133 96 66 43 26 12 1 8 16

ROI ¼ return on investment. * Based on recruitment rate of 200 patients per site per year and additional 1 y for data analysis. y Total trial cost ¼ fixed þ variable costs þ opportunity cost. z ENBS ¼ the difference between EVSI and total trial cost. x ROI ¼ ENBS/total cost.

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909

8

910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

Table 6 e Sensitivity analysis of assumptions effect on the cost-effectiveness and value of additional analyses results. Assumption NPWT price Baseline 25% increase 25% reduction Willingness-to-pay threshold/QALY Baseline 50% increase 50% reduction RR Baseline (Masden et al. [25] and pilot trial) [36]. Masden et al. [25] alone Pilot trial alone [36]. Technology lifetime Baseline 50% increase 50% reduction

Estimate

Incremental net benefit

Value of information

Optimal sample size/arm

AUD175 AUD220 AUD130

AUD70 AUD30 AUD125

AUD2.7 million AUD3.2 million AUD2.0 million

200 250 150

AUD50,000 AUD75,000 AUD25,000

AUD70 AUD125 AUD30

AUD2.7 million AUD2.8 million AUD2.7 million

200 250 200

0.73

AUD70

AUD2.7 million

200

0.5 0.81

AUD200 AUD -35

AUD3.2 million AUD3.8 million

250 300

10 y 15 y 5y

AUD70 AUD70 AUD70

AUD2.7 million AUD3.4 million AUD1.6 million

200 300 100

QALY ¼ quality adjusted life year.

economic evaluations. Nevertheless, this should not pose a problem because when evidence is scarce, information could be sought from various sources such as pilot trials, observational studies, and expert opinion [55]. Ideally, this should be also accompanied by an appropriate value of information analysis to inform whether that evidence is sufficient to guide decisions or if additional research is required. For instance, Soares et al. [55] conducted cost-effectiveness and value of information analyses on NPWT in patients with severe pressure ulcers. They demonstrated how combining information from the existing evidence with a pilot trial and elicited expert views resulted in a better informed decision compared with using a single source of evidence when information is scarce. Moreover, they used value of information analysis to optimize future trial design [55]. In this article, we populated our model with the best available evidence in the literature combined with the results of our pilot trial and expert opinion when necessary. As expected for any economic evaluation, the results of our cost-effectiveness analysis are dependent on the assumptions made for the model structure and input parameters. We used hydrocolloid dressing, which is the standard of care for this procedure in Australia, as the comparator in our analysis; however, there are other types of surgical dressings in clinical practice at various prices. Importantly, it is essential not to limit the comparison to the unit prices of the products but also to consider the efficacy and overall cost of use. For instance, NPWT was less costly than saline soaked gauze (US $14,546 compared with US $23,465) in healing pressure ulcers because wounds healed 61% faster with NPWT compared with the standard gauze and saline [60]. Furthermore, our model focused on SSI as an outcome and did not include other outcomes such as healing rate or other wound complications. However, the model did not include healing as an outcome because, unlike chronic wounds, most clean incision wounds will completely heal in a relatively short time [21]. Additionally, compared with other wound complications expected

with CS (e.g., seroma), SSIs are more associated with mortality, morbidity, and cost. In addition, our model was probabilistic and Monte Carlo sampling allowed for simultaneous characterization of uncertainty in all model parameters. Finally, we tested the effect of various assumptions made to the value of information analysis on the results. In the sensitivity analysis presented, the optimal sample size remained between 100 and 300 patients in most of the scenarios.

5.

Conclusions

Based on the best available evidence, NPWT appears costeffective compared with standard dressing in preventing SSI in obese women undergoing CS. But, there is high uncertainty surrounding a decision to implement this technology and further research to explore the relative effect of NPWT in this population would be worthwhile before implementation.

Uncited figure Figure 2.

Acknowledgments H.W.T. is supported by a National Health and Medical Q6 Research Council PhD scholarship through the Centre for Research Excellence in Nursing Interventions for Hospitalised Patients. Authors’ contributions: H.W.T., L.G.G., and P.A.S. performed the economic analysis. W.C. and B.M.G. provided the clinical data. All authors contributed substantially to the preparation of the article.

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104

Disclosure [22]

The authors declare no conflict of interest.

Q11

Q10

references

[23]

[1] Abrams B, Heggeseth B, Rehkopf D, Davis E. Parity and body mass index in US women: a prospective 25-year study. Obesity 2013;21:1514. [2] Vahratian A. Prevalence of overweight and obesity among women of childbearing age: results from the 2002 National Survey of Family Growth. Matern Child Health J 2009;13:268. [3] Department of Health, Health Survey for England 2012. [4] Australian Bureau of Statistics. Australian Health Survey: updated results, 2011-2012. [5] Heslehurst N, Simpson H, Ells LJ, et al. The impact of maternal BMI status on pregnancy outcomes with immediate short-term obstetric resource implications: a meta-analysis. Obes Rev 2008;9:635. [6] Chu SY, Kim SY, Schmid CH, et al. Maternal obesity and risk of cesarean delivery: a meta-analysis. Obes Rev 2007;8:385. [7] Kominiarek MA, Vanveldhuisen P, Hibbard J, et al. The maternal body mass index: a strong association with delivery route. Am J Obstet Gynecol 2010;203:e1. [8] Callaway LK, Prins JB, Chang AM, McIntyre HD. The prevalence and impact of overweight and obesity in an Australian obstetric population. Med J Aust 2006;184:56. [9] Anderson V, Chaboyer W, Gillespie B. The relationship between obesity and surgical site infections in women undergoing caesarean sections: an integrative review. Midwifery 2013;29:1331. [10] Centers for Disease Control and Prevention, Surgical site infection (SSI). [11] Johnson A, Young D, Reilly J. Caesarean section surgical site infection surveillance. J Hosp Infect 2006;64:30. [12] Urban JA. Cost analysis of surgical site infections. Surg Infect 2006;7(Suppl 1):S19. [13] Wilson J, Wloch C, Saei A, et al. Inter-hospital comparison of rates of surgical site infection following caesarean section delivery: evaluation of a multicentre surveillance study. J Hosp Infect 2013;84:44. [14] Graves N, Halton K, Doidge S, Clements A, Lairson D, Whitby M. Who bears the cost of healthcare-acquired surgical site infection? J Hosp Infect 2008;69:274. [15] Schweizer ML, Cullen JJ, Perencevich EN, Vaughan Sarrazin MS. Costs associated with surgical site infections in veterans affairs hospitals. JAMA Surg; 2014. [16] Ahmed SR, Ellah MA, Mohamed OA, Eid HM. Prepregnancy obesity and pregnancy outcome. Int J Health Sci 2009;3:203. [17] Alanis MC, Villers MS, Law TL, Steadman EM, Robinson CJ. Complications of cesarean delivery in the massively obese parturient. Am J Obstet Gynecol 2010;203:e1. [18] Gillespie B, Finigan T, Kerr D, Lonie G, Chaboyer W. Endusers’ assessment of prophylactic negative pressure wound therapy products. Wound Pract Res 2013;21:74. [19] Othman D. Negative pressure wound therapy literature review of efficacy, cost effectiveness, and impact on patients’ quality of life in chronic wound management and its implementation in the United Kingdom. Plast Surg Int 2012;2012:374398. [20] Suissa D, Danino A, Nikolis A. Negative-pressure therapy versus standard wound care: a meta-analysis of randomized trials. Plast Reconstr Surg 2011;128:498e. [21] Webster J, Scuffham P, Sherriff KL, Stankiewicz M, Chaboyer WP. Negative pressure wound therapy for skin

[24]

[25]

[26]

[27]

[28]

[29]

[30] [31] [32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

9

grafts and surgical wounds healing by primary intention. Cochrane Database Syst Rev 2012;4:CD009261. Ingargiola MJ, Daniali LN, Lee ES. Does the application of incisional negative pressure therapy to high-risk wounds prevent surgical site complications? A systematic review. Eplasty 2013;13:e49. Webster J, Scuffham P, Stankiewicz M, Chaboyer WP. Negative pressure wound therapy for skin grafts and surgical wounds healing by primary intention. Cochrane Database Syst Rev 2014;10:CD009261. Howell R, Hadley S, Strauss E, Pelham F. Blister formation with negative pressure dressings after total knee arthroplasty. Curr Orthopaedic Pract; 2011:176. Masden D, Goldstein J, Endara M, Xu K, Steinberg J, Attinger C. Negative pressure wound therapy for at-risk surgical closures in patients with multiple comorbidities: a prospective randomized controlled study. Ann Surg 2012;255: 1043. Stannard JP, Volgas DA, McGwin G 3rd, et al. Incisional negative pressure wound therapy after high-risk lower extremity fractures. J Orthop Trauma 2012;26:37. Tuffaha HW, Gordon LG, Scuffham PA. Value of information analysis in healthcare: a review of principles and applications. J Med Econ 2014;17:377. Steuten L, van de Wetering G, Groothuis-Oudshoorn K, Retel V. A systematic and critical review of the evolving methods and applications of value of information in academia and practice. Pharmacoeconomics 2013;31:25. Tuffaha HW, Reynolds H, Gordon LG, Rickard CM, Scuffham PA. Value of information analysis optimizing future trial design from a pilot study on catheter securement devices. Clin Trials 2014;11:648. Willan AR, Pinto EM. The value of information and optimal clinical trial design. Stat Med 2005;24:1791. Claxton K, Posnett J. An economic approach to clinical trial design and research priority-setting. Health Econ 1996;5:513. Briggs AH, Goeree R, Blackhouse G, O’Brien BJ. Probabilistic analysis of cost-effectiveness models: choosing between treatment strategies for gastroesophageal reflux disease. Med Decis Making 2002;22:290. Stinnett AA, Mullahy J. Net health benefits: a new framework for the analysis of uncertainty in cost-effectiveness analysis. Med Decis Mak 1998;18:S68. Harris AH, Hill SR, Chin G, Li JJ, Walkom E. The role of value for money in public insurance coverage decisions for drugs in Australia: a retrospective analysis 1994-2004. Med Decis Making 2008;28:713. Opoien HK, Valbo A, Grinde-Andersen A, Walberg M. Postcesarean surgical site infections according to CDC standards: rates and risk factors. A prospective cohort study. Acta Obstet Gynecol Scand 2007;86:1097. Chaboyer W, Anderson V, Webster J, Sneddon A, Thalib L, Gillespie BM. Negative pressure wound therapy on surgical site infections in women undergoing elective caesarean sections: a pilot RCT. Healthcare 2014;2:417. Henman K, Gordon CL, Gardiner T, et al. Surgical site infections following caesarean section at Royal Darwin Hospital, Northern Territory. Healthc Infect 2012;17:47. Astagneau P, Rioux C, Golliot F, Brucker G. Group INS. Morbidity and mortality associated with surgical site infections: results from the 1997-1999 INCISO surveillance. J Hosp Infect 2001;48:267. Kirkland KB, Briggs JP, Trivette SL, Wilkinson WE, Sexton DJ. The impact of surgical-site infections in the 1990s: attributable mortality, excess length of hospitalization, and extra costs. Infect Control Hosp Epidemiol 1999;20:725. Clemens S, Begum N, Harper C, Whitty JA, Scuffham PA. A comparison of EQ-5D-3L population norms in Queensland,

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169

10

1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228

[41]

[42]

[43] [44]

[45]

[46]

[47]

[48]

[49]

[50]

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

Australia, estimated using utility value sets from Australia, the UK and USA. Qual Life Res 2014;23:2375. Lipsky BA, Moran GJ, Napolitano LM, Vo L, Nicholson S, Kim M. A prospective, multicenter, observational study of complicated skin and soft tissue infections in hospitalized patients: clinical characteristics, medical treatment, and outcomes. BMC Infect Dis 2012;12:227. Graves N, Halton K, Curtis M, et al. Costs of surgical site infections that appear after hospital discharge. Emerg Infect Dis 2006;12:831. Australian Institute of Health and Welfare, Australian refined diagnosis-related groups (AR-DRG) 2011. 2010. Spiegelhalter DJ, Myles JP, Jones DR, Abrams KR. Bayesian methods in health technology assessment: a review. Health Technol Assess 2000;4:1. Tuffaha HW, Rickard CM, Webster J, et al. Cost-effectiveness analysis of clinically indicated versus routine replacement of peripheral intravenous catheters. Appl Health Econ Health Policy 2014;12:51. Shemilt I,. CCEMG-EPPI-Centre Cost Converter; version 1.4. The Campbell and Cochrane Economics Methods Group (CCEMG) and the Evidence for Policy and Practice Information and Coordinating Centre (EPPI-Centre). Available from: http://eppi.ioe.ac.uk/costconversion/default. aspx. Accessed July 2014. Ades AE, Lu G, Claxton K. Expected value of sample information calculations in medical decision modeling. Med Decis Mak 2004;24:207. Brennan A, Kharroubi S, O’Hagan A, Chilcott J. Calculating partial expected value of perfect information via Monte Carlo sampling algorithms. Med Decis Mak 2007;27:448. Claxton K. The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies. J Health Econ 1999;18:341. Claxton KP, Sculpher MJ. Using value of information analysis to prioritise health research: some lessons from recent UK experience. Pharmacoeconomics 2006;24:1055.

[51] Groot Koerkamp B, Myriam Hunink MG, Stijnen T, Weinstein MC. Identifying key parameters in costeffectiveness analysis using value of information: a comparison of methods. Health Econ 2006;15:383. [52] Australian Institute of Health and Welfare, Australia’s mothers and babies 2010. [53] Eckermann S, Karnon J, Willan AR. The value of value of information best informing research design and prioritization using current methods. Pharmacoeconomics 2010;28:699. [54] McKenna C, Claxton K. Addressing adoption and research design decisions simultaneously: the role of value of sample information analysis. Med Decis Mak 2011;31:853. [55] Soares MO, Claxton K, Cullum N, et al. Methods to assess cost-effectiveness and value of further research when data are sparse: negative-pressure wound therapy for severe pressure ulcers. Med Decis Making; 2012. [56] Willan AR, Goeree R, Boutis K. Value of information methods for planning and analyzing clinical studies optimize decision making and research planning. J Clin Epidemiol 2012;65:870. [57] de Leon JM, Barnes S, Nagel M, Fudge M, Lucius A, Garcia B. Cost-effectiveness of negative pressure wound therapy for postsurgical patients in long-term acute care. Adv Skin Wound Care 2009;22:122. [58] Le Franc B, Sellal O, Grimandi G, Duteille F Cost-effectiveness analysis of vacuum-assisted closure in the surgical wound bed preparation of soft tissue injuries. Ann Chir Plast Esthet 2010;55:195. [59] Dowsett C, Davis L, Henderson V, Searle R. The economic benefits of negative pressure wound therapy in communitybased wound care in the NHS. Int Wound J 2012;9:544. [60] Philbeck TE Jr, Whittington KT, Millsap MH, Briones RB, Wight DG, Schroeder WJ. The clinical and cost effectiveness of externally applied negative pressure wound therapy in the treatment of wounds in home healthcare Medicare patients. Ostomy Wound Manage 1999;45:41.

1229 1230 1231 1232 1233 5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298

11

j o u r n a l o f s u r g i c a l r e s e a r c h x x x ( 2 0 1 5 ) 1 e1 1

1299 1300 1301 Q7 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363

3. Expected value of sample information

Appendix

EVSI is the difference between the expected value of a decision made after collecting data (D) on the parameter of interest and the expected NB with current information [47].

Appendix 1.

The NB for an intervention i informed by the set of input parameters q was calculated as follows: NB ði; qÞ ¼ l Effect ði; qÞ  Cost ði; qÞ

(1)

1. Expected value of perfect information The EVPI is the difference between the expected NB of a decision with perfect information and the decision made based on current information. [49]. EVPI ¼ Eq Maxi NBði; qÞ  Maxi Eq NBði; qÞ

(2)

1. Assigned probability distributions to the input parameters in the mode as summarized in Table 2 in the main text. 2. Sampled random values k times (e.g., k ¼ 10,000) from the distributions described previously for each intervention. 3. Calculated the mean NB for each intervention across all simulations and identified the preferred baseline decision that is, the intervention with the maximum expected mean NB (maxi Eq NBði; qÞ). 4. Calculated the NB for each intervention and identified the optimal intervention at each simulation. 5. Averaged the NBs from the identified optimal interventions in step 4 (Eq maxi NBði; qÞ). 6. EVPI per patient is the difference between the average NBs in steps 5 and 3.

2. Expected value of perfect parameter information Because the model used was linear and assuming no correlation between input parameters, the same one-level Monte Carlo simulation technique described previously was used to calculate the EVPPI; the sampling would be only from the distribution of the parameter(s) of interest qI , whereas the other parameters qC were fixed at their prior means [51]. EVPPIqI ¼ EqI maxi NBði; qI ; EðqC ÞÞ  maxi Eq NBði; qÞ

(3)

Steps 1e3 are as described in EVPI algorithm detailed previously. Steps 4e9 are as follows: 4. Sample qI once from its prior distribution (one-level simulation). 5. Fix qI at their sampled values, and fix the remaining uncertain parameters qC at their prior mean value. 6. Calculate the average NB of each intervention given these parameter values. 7. Identify the intervention that has the highest estimated expected NB given the sampled value for the parameters of interest (qIK ). 8. Repeat steps 4e7 k times (e.g., k ¼ 10,000), and calculate the average NB of the preferred interventions identified in step 7. 9. EVPPI is the difference between the average NBs in steps 8 and 3.

EVSIn ¼ ED maxi NBði; EðqI jDÞ; EðqC ÞÞ  maxi Eq NBði; qI ; qC Þ

(4)

Given the linearity of the model, calculating the EVSI required the same one-level Monte Carlo sampling; however, the sampling would be from the posterior distribution of the parameter of interest obtained using Bayesian updating [47]. To estimate the EVSI for the RR of NPWT compared with that of standard dressing, we assumed that parameters qNP and qSD represent the probability of SSI with NPWT and standard dressing, respectively. We followed the algorithm adapted from the algorithm reported in Ades et al. [47]. Steps 0e3 are as described in EVPI algorithm mentioned previously. Steps 4e9 are given as follows: 4. Simulate the variety of possible results of proposed data collection by the following steps. 4.1. Draw a sample from the prior distribution of the RR. The logRR w normal (m0 , s0 ) wherem0 is logRR in the metaanalysis and s0 is its variance. 4.2. Draw a sample baseline parameter qSD from its prior distribution: qSD w beta (a,b), where a is the number of patients who developed SSI and b is the number of patients who did not develop SSI from the combined data of the pilot trial and Masden et al. 4.3. Transform back to obtain an implied prior for qNP : qNP ¼ qSD expðlog RRÞ 5. 5.1 Draw a sample sufficient statistic D, in this case a binominal numerator, for each arm in the future trial with size n, assuming equal size arms: rSD wbinomial ðqSD ; nÞandrNP wbinomial ðqNP ; nÞ 5.2 Convert the sufficient statistics to a mean and variance using the normal approximation:    mD ¼ log rNP n rSD n ; 

1  sD ¼ ðn  rSD Þ rSD n þ ðn  rNP Þ rNP n 6. Update the prior with the new simulated data to obtain parameters of the posterior distribution: logRRjDwnormal m0 s0 þ mD ; sD



ðs0 þ sD Þ; s0 þ sD

7. Because the model is linear, we sampled from the expected value of the updated distribution in step 6 and the mean values for qC and identified the intervention with the highest expected NB. 8. Repeat steps 4e7 for 10,000 times, and calculate the average NB of the preferred interventions identified in step 7. 9. The EVSI is the difference between the average NBs in steps 8 and 3.

5.2.0 DTD
YJSRE13128_proof
28 February 2015
1:25 pm
ce

1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428