Reference case. For our analysis, the reference case was an adult diagnosed with complicated appendicitis with confirmatory abdominal imaging. The patient was > 18 years of age without comorbidities that would substantially increase their risk of complications from laparoscopic appendectomy and image-guided drainage. Complicated appendicitis was defined as appendiceal inflammation with the presence of appendiceal abscess, cellulitis, or extraluminal air on initial abdominal computed tomographic images.
Treatment Strategies. We compared the cost-effectiveness of the three following treatment strategies from a health-care payer’s perspective in Japan (health insurers, and the government): (1) operative management with elective ELA, (2) initial nonoperative management with interval laparoscopic appendectomy (ILA) at 2 months, and (3) nonoperative management without ILA. Nonoperative management entailed hospitalization for 5 days with intravenous cefmetazole as a 3-day course of antibiotics. Computed tomography-guided percutaneous abscess drainage was performed if necessary. All nonoperative treatment failures regardless of the specific indication (e.g., failure to improve, worsening vital signs or laboratory parameters, provider or patient preference) required delayed laparoscopic appendectomy in the same hospitalization. Failures occurring after discharge were considered recurrent appendicitis, and patients with recurrent appendicitis underwent ELA.
Decision model. For the analysis, we used decision analytical modeling using a Markov model to simulate costs, health outcomes, and cost savings while comparing the three treatment strategies (Fig. 1). Given that patients’ health states generally return to baseline within 1 year after acute appendicitis, we chose to condense our model by applying the total long-term risk of recurrent appendicitis after nonoperative management into a single year [10]. therefore, we did not include background mortality in the simulation. Each cycle was defined as 1 month in length. The decision model was constructed and analyzed using R, version 3.5.0 with heemod package(R Foundation for Statistical Computing, Vienna, Austria).
Probabilities
Probabilities of clinical events were abstracted from a literature review (Table 1) [4] [11] [12]. The literature review was performed using the PubMed database, using the terms, "Acute appendicitis," "Nonoperative," "Conservative," "Nonsurgical," "Appendectomy," "Complicated," "Abscess,” ”Perforated," and "Phlegmon." These terms and their combinations were also searched as text words. The search was performed on November 2018, and English language restriction was applied. Because there was limited evidence of the efficacy of laparoscopic appendectomy compared with nonoperative management in treating complicated appendicitis, the inclusion criteria were randomized clinical trials and meta-analyses comparing nonoperative management and operative management (which includes laparoscopic appendectomy) in complicated appendicitis. Exclusion criteria were narrative reviews, studies without control groups, case reports, case series studies, and studies involving pediatric patients.
Costs. Costs were estimated from a health-care payer’s perspective; therefore, only direct medical costs were included. The cost of laparoscopic appendectomy is an unexplored field, and careful attention should be paid to the costing methodology because there is no gold standard. Therefore, for cost analysis, we used a micro-costing method in which the actual monetary health care costs are categorized within the main category: Diagnostic procedures, drugs, ward care, and operating room cost [13] [14]. Data for health care costs were based on the diagnosis-procedure combination/per-diem payment system and fee-for-service, and specific material expenses between 01 April 2011 and 31 March 2018 were retrospectively obtained from the electronic database of the Tochigi Medical Center (Table 2). Costs of perioperative complications were estimated by the increase in average hospitalization cost for complicated appendicitis with complication or comorbidity based on data from our hospital records. We assumed that the costs for outpatient follow-up after hospitalization were equivalent between treatment groups, so these costs were not included in this analysis. Costs are expressed based on US dollars, 2018.
Health-related utility. The primary measures of effectiveness in the present analysis were quality- adjusted life years (QALY) gained. To estimate total QALYs in the Markov model, QALYs were calculated by multiplying the health care-related quality of life (HRQoL) score of a disease state by the duration of time a patient spent in that disease state. We obtained the HRQoL factors from our literature review, and data are shown in Table 3 [10] [15] [16] [17]. Because there were few quality-of-life estimates in the literature for the health states of appendicitis [18] [19] [20]. we used the method proposed by Wu et al, in which the utility of undergoing various treatments for appendicitis is estimated by multiplying established utilities by the average duration of hospitalization and recovery from complicated appendicitis associated with each therapeutic strategy [15] [16] [17]. Additionally, we utilized 1-time decreases in QALY for unplanned emergency room visit with readmission and percutaneous drainage of abdominal abscess [17]. For example, given that the utility of patients hospitalized for nonsurgical treatment was assumed to be reduced to 0.98 for the duration of the hospitalization and the mean length of stay for these patients was assumed to be 5 days, and One-time QALY reduction for unplanned emergency visit was subtracted from total QALYs, the utility for that a month cycle (28 d) was calculated as ([0.98*5]+ [23*1])/28-0.005= 0.991 (for patients with antibiotic therapy).
Cost-effectiveness analysis
Cost-effectiveness was evaluated using the incremental cost-effectiveness ratio (ICER). In this analysis, we defined the willingness-to-pay (WTP) threshold based on the criterion of the World Health Organization that states that an intervention is considered cost-effective if the ICER for QALY is 1–3-fold the gross domestic product per capita [21], which supports the $107,690 per QALY threshold for cost effectiveness used in this study. Based on this standard, we defined the preferred strategy as the strategy that produced the greater utility without exceeding a threshold of $107,690 per QALY. A strategy is considered dominant if it is both less costly and more effective than the reference strategy. Operative management, the current standard of care, was used as the reference group for all comparisons.
Sensitivity analysis
We performed several sensitivity analyses to evaluate the uncertainty and robustness of the model. For these sensitivity analyses, we selected the parameters that covered all potential areas of uncertainty, such as the probabilities, clinical costs, and health-related utility values. One-way sensitivity analyses assessed the effects of varying key model parameters on the ICER. The variation ranges were established based on the analyzed studies. For costs, we allowed values to vary by ± 50% of the index value; variations in sensitivity analysis results are listed in Tables 1–3. We also performed a probabilistic sensitivity analysis to assess the impact of sensitivity on the model parameters using a Monte Carlo simulation with 1000 samples. For the probabilistic sensitivity analysis, all model variables (probabilities, costs, utilities) were set as static with triangular frequency distributions. Additionally, a threshold analysis was performed to determine the cost-effective price of each treatment strategy.
Ethics
This study was approved by the local Ethics Committee on December 3rd, 2018 (number 2018110501).