Cost-effectiveness of Imaging Protocols for Suspected Appendicitis

We simulated a hypothetical cohort of children presenting to an ED with suspected appendicitis using a decision-analytic Markov cohort model created in TreeAge (TreeAge Pro Healthcare; TreeAge Software Inc, Williamstown, MA). We compared the following imaging strategies (Fig 1): CT only (strategy A); MRI only (strategy B); or a staged approach using ultrasound followed by CT or MRI (1) after all ultrasounds that are either negative or do not visualize the appendix (strategies C and F), (2) after all ultrasounds that do not visualize the appendix (strategies D and G), or (3) only after ultrasounds that do not visualize the appendix if the ultrasound has secondary signs of inflammation in the RLQ (strategies E and H). Lastly, we modeled discharge and surgery for all patients without any imaging (strategies I and J, respectively) to evaluate the test and test-treatment threshold of imaging. Strategies represented existing described protocols and the existing literature.^7,20,22 

We estimated the costs of imaging and surgical procedures, and the costs and health effects attributable to false-negative and positive diagnoses, including increased risk of perforation and negative appendectomy, respectively. We assumed patients who had imaging tests that were falsely interpreted as negative returned to the ED, at which point they would be diagnosed with appendicitis and have an increased perforation risk. A simplified version of 1 arm of the decision tree (strategy E) is shown in Fig 2. The full decision tree is provided (Supplemental Figs 4 and 5). The study was deemed exempt by the Seattle Children’s Hospital Institutional Review Board.

We estimated input parameters including imaging test characteristics and health outcomes of children with suspected acute appendicitis from published studies (Table 1). We used meta-analyses describing imaging test characteristics when available.^12,23 

For ultrasound test characteristics, we included published studies that compared positive or negative ultrasound results for pediatric patients with suspected appendicitis to positive or negative rates of appendicitis at surgery or follow-up. Test characteristics of ultrasounds that visualize the appendix were estimated from published studies that excluded nonvisualized appendices.^{8,9,16–18,24–27,30–33}  Test characteristics of the presence of secondary signs of inflammation in ultrasounds that do not visualize the appendix were estimated from studies that separately reported the test characteristics when the appendix is not visualized, with and without secondary signs,9,16,24,27,30-32 assuming that patients with secondary signs had appendicitis and those without secondary signs did not have appendicitis. We used a weighted average from these studies for sensitivity and specificity point estimates.

We estimated visualization rates for patients with and without appendicitis separately using weighted averages from published data.^{8,9,16,18,24–27,30–33}  None of the appendix visualization studies differentiated between ruptured and nonruptured appendices.

Mortality from a negative appendectomy was assumed to be equal to mortality for pediatric patients with uncomplicated appendicitis.

Health-related quality of life associated with perforated and uncomplicated appendicitis was estimated by using the Pediatric Quality of Life Inventory (PedsQL), which was administered to patients hospitalized at our freestanding children’s hospital from June 2009 to April 2018. The PedsQL, a validated standardized questionnaire, produces scores on 4 subscales: physical, emotional, social, and school functioning. We used published regression equations to estimate utility values from the PedsQL subscales.⁴⁴  We identified patients using International Classification of Diseases, Ninth Revision, Clinical Modification and International Classification of Diseases, 10th Revision, Clinical Modification codes of 540.9 and K35.80 or K35.89 for acute appendicitis and 540.0 and K35.2 or 540.1 and K35.3 for perforated appendicitis.²¹  We assumed the utility value for negative appendectomy would be the same as for uncomplicated acute appendicitis. We used published PedsQL data for ED patients with minor injuries to estimate the utility values for ED patients presenting with abdominal pain but no appendicitis, defined as the “well state.”^42,43 

Inpatient costs for uncomplicated and perforated appendicitis were estimated by using the 2016 Kids’ Inpatient Database (KID), a national database of hospital charge data. We defined patients with uncomplicated and perforated appendicitis using International Classification of Diseases, 10th Revision, Clinical Modification codes of K35.80 or K35.89 and K35.2 or K35.3, respectively. We assumed that the cost of negative appendectomies was the same as that of uncomplicated appendicitis. We also included professional fees in total costs (not included in KID cost data) by including professional charges based on assumed Current Procedural Terminology (CPT) codes charged by providers²¹  and average length of stay in the KID for uncomplicated and perforated appendicitis. We estimated imaging and ED visits costs using 2017 Medicare hospital fees⁴⁰  and professional fees by CPT code.⁴¹  Total cost estimates are included in Table 1, and the components are shown in Supplemental Table 3. All costs were adjusted to 2017 US dollars by using the Personal Health Care Expenditure deflator.⁴⁵ 

A cost-effectiveness analysis was conducted from the payer perspective over a 1-year time horizon.⁴⁶  Our primary effectiveness measure was the quality-adjusted life-year (QALY), a composite measure of morbidity and mortality.⁴⁶  QALYs for each strategy were calculated as the number of years spent living a health state (ie, appendicitis or appendicitis with perforation) multiplied by the utility value for that health state (Table 1). Because appendicitis and ED visits are temporary health states, the patient was assumed to experience decreased utility for 1 month then return to a state of perfect health for the remaining 11 months of the year. We ranked each of the 10 alternative screening approaches according to increasing cost. Approaches that were more costly but less effective than an alternative approach were considered dominated and excluded from further calculations.⁴⁶  The remaining approaches were reranked according to increasing costs. The least expensive and least effective screening approach was the baseline for comparison. We then calculated the net costs, the cost of the strategy minus the cost of the next-lowest–ranking strategy, and net QALYs (the effectiveness of the strategy minus the effectiveness of the next-lowest–ranking strategy) for each screening alternative. Our cost-effectiveness outcome was the incremental cost-effectiveness ratio (ICER) for each alternative, or net costs divided by net QALYs.⁴⁷ 

We conducted 1- and 2-way sensitivity analyses to assess how robust our results were to inputs. Ranges are reported in Table 1. In our 2-way sensitivity analysis, we varied the pretest probability of appendicitis and proportion of ultrasounds that visualize the appendix from 0% to 100%. The latter was varied at a fixed ratio for patients with and without appendicitis. We assessed thresholds at which the most cost-effective approach changes. This allowed us to create a tailored imaging approach based on a patient’s pretest probability of appendicitis and a hospital’s appendix visualization rate.

Results are summarized in Table 2. Strategy E is the least costly strategy, with a cost of $4815 and 0.9969 QALYs. In this strategy, all patients with suspected appendicitis undergo an initial ultrasound. If the ultrasound does not visualize the appendix but has secondary signs of inflammation, then the patient undergoes a CT. If there are no secondary signs of inflammation on ultrasound, then the patient is discharged. All but 3 strategies were dominated, with lower effectiveness and higher cost than their next least expensive comparators being observed (Table 2, Supplemental Table 4). Strategies B (MRI only) and H (staged ultrasound with MRI only if ultrasound does not visualize the appendix but has secondary signs of inflammation) provide small QALY gains over strategy E, but the ICER for each is significantly higher than standard willingness-to-pay thresholds of $100 000 to $150 000 per QALY gained.⁴⁶ 

The 1-way sensitivity analysis demonstrated that ICERs were most sensitive to the prevalence of appendicitis, the cost of appendicitis treatment, and the specificity of ultrasounds with appendix visualization (Supplemental Fig 6). The other inputs had little impact on ICERs, meaning that within the range of the input that was simulated, the ICER did not vary substantially.

A 2-way sensitivity analysis (Fig 3) demonstrated little effect of the proportion of visualized ultrasounds on optimal strategies even when visualization is low. Strategy I (discharge all) is the optimal strategy when the pretest probability of having appendicitis is <∼16%. Strategy E is optimal when the pretest probability is between 16% and 67%. With higher pretest probability (67%–95%), strategy D is optimal (all nonvisualized appendices on ultrasound are followed by CT). When the pretest probability is >95%, then strategy J (surgery without imaging) is optimal.

Because the absence of secondary signs of inflammation in the RLQ when the ultrasound does not visualize the appendix has a lower sensitivity than a negative ultrasound result when the appendix is seen, it is only cost-effective to forego a CT with a nonvisualized ultrasound if there are no secondary signs of RLQ inflammation and the patient has a lower risk of appendicitis. To maximize cost-effectiveness, when a patient has a higher risk of appendicitis, all ultrasounds that do not visualize the appendix should be followed by CT. The sensitivity of an ultrasound that visualizes the appendix is high enough that a CT after a negative ultrasound result is only cost-effective when the proportion of nonvisualized ultrasounds is high. Furthermore, we found that if the patient has a low or high pretest probability of appendicitis, it is cost-effective for patients to forego imaging entirely.

In this study, we provide guidance to clinicians regarding the most cost-effective imaging strategies for pediatric patients with suspected appendicitis, with particular consideration on risk stratifying a patient’s pretest probability of appendicitis and a hospital’s proportion of ultrasounds that are unable to visualize the appendix. The most cost-effective strategy is highly dependent on a patient’s risk stratification. At a prevalence <16% and >95%, patients can forego imaging entirely. For patients with a risk of appendicitis between 16% and 67%, it is cost-effective to perform an initial ultrasound and forego additional imaging if the ultrasound does not visualize the appendix but shows no secondary signs of inflammation. When the pretest probability of appendicitis is >67%, it is cost-effective to follow-up all nonvisualized ultrasounds with a CT even without secondary signs of inflammation on ultrasound.

We found that even when the appendix is rarely visualized, ultrasound as the initial imaging modality for intermediate-risk patients is cost-effective. This is important because most quality-improvement initiatives that examined the ultrasound-first approach have been conducted in children’s hospitals, but most pediatric patients are seen in general EDs,⁴⁸  where the proportion of ultrasounds that visualize the appendix may be lower. Our results suggest all hospitals should adopt risk stratification followed by eventual ultrasound in patients with intermediate probability of appendicitis. It is important to note that our results were sensitive to changes in the specificity of ultrasound with visualization of the appendix; when ultrasound is not read by a pediatric radiologist, it may have a lower specificity, which would impact the cost-effectiveness of ultrasound-first approaches.

The optimal imaging strategy is highly dependent on a patient’s pretest probability of appendicitis; therefore, imaging strategies should be tailored on the basis of a provider’s assessment of an individual patient’s risk of appendicitis. Providers can estimate a patient’s appendicitis risk using ancillary tests or scores such as the Pediatric Appendicitis Risk Calculator⁴⁹  or Pediatric Appendicitis Score.⁵⁰  Our findings are consistent with previous studies and European guideline recommendations.^18,19,51  The risk thresholds for imaging we identified are similar to those derived from risk prediction tools. The Pediatric Appendicitis Risk Calculator scoring tool identifies patients with medium risk of appendicitis from 15% to 85%, similar to our model’s recommendation of imaging at a risk level of 16% to 95%.⁴⁹  Furthermore, our results align with previous findings that patients with a low likelihood of appendicitis can be discharged with an ultrasound that does not visualize the appendix but has no secondary signs of inflammation given the high negative predictive value of an ultrasound without secondary signs in low-risk clinical settings.^9,32  It is important to note that many of these baseline estimates of pretest probability are from studies at children’s hospitals.^18,50  The advantage of using our model to establish test thresholds is that it can accurately estimate rare outcomes, such as missed diagnosis leading to increased perforation risk, and identify precise pretest probabilities at which different imaging approaches are most cost-effective.

Given the high cost of surgery and hospitalization relative to imaging in the ED, we found that it is only cost-effective to forego imaging when there is a high risk for appendicitis. Our own personal experience is that there is much variation in the approaches that providers use with high-risk patients. This exemplifies why decision models using virtual cohorts are helpful for informing clinical practice.

Many hospitals have adopted quality-improvement initiatives aimed at reducing CT use for suspected appendicitis to reduce exposure to ionizing radiation. Given high uncertainty about long-term effects of radiation levels seen with a single abdominal CT,⁵²  we did not model potential future cases of radiation-induced cancer. Importantly, the CT-only protocol was not cost-effective even without considering radiation. Therefore, an ultrasound-first protocol would not only reduce patients’ exposure to ionizing radiation but also improve value.

Because the appendicitis treatment and recovery period is brief, there were only minor differences in QALYs. It remains important to consider the effect QALYs have on our results because ICERs identify dominant approaches (both more effective and less costly). A cost-only analysis would mask these dominant scenarios. With a higher sensitivity and specificity than CT, MRI-based imaging strategies have minimally higher QALY gains than do CT-based strategies. Because of the higher cost of MRI, the ICERs for MRI-based strategies are significantly higher than accepted willingness-to-pay thresholds. New short-duration, appendix-specific MRI protocols could make MRI more cost-effective if those protocols reduce costs substantially.¹³ 

There has been recent interest in the treatment of assumed uncomplicated appendicitis with antibiotics without surgery.^53,54  Wu et al⁴³  demonstrated that this approach can be cost-effective. Because this approach costs less than operative management,⁵⁴  it may affect the cost-effectiveness of diagnosis. Recent studies have suggested some cases of uncomplicated appendicitis may resolve spontaneously,⁵³  suggesting there may be a risk of overtreating mild appendicitis detected by ultrasound or CT. Furthermore, some patients with false-negative diagnosis of appendicitis may have spontaneous resolution of their mild appendicitis.

This analysis was subject to limitations. We did not account for all possible costs and consequences related to imaging and outcomes. Perforated appendicitis can lead to abdominal strictures and scarring, leading to obstruction and potential future surgeries. Imaging can lead to incidental findings with unclear clinical significance, which may necessitate further imaging and invasive procedures. We did not account for sedation when assessing MRI costs and morbidity. Many input parameter estimates are based on studies performed at children’s hospitals. As is common in decision-analytic models, we made multiple modeling assumptions. However, sensitivity analyses provide robustness checks on our assumptions and results. Finally, we could not extract imaging costs from total hospital costs; therefore, costs for patients with appendicitis include 2 imaging modalities. However, imaging represents a fraction of total treatment costs, and this change is unlikely to affect our results.

Accurate diagnosis of appendicitis can be challenging, with consequences for both false-positive and false-negative diagnoses. It is important to minimize potential negative consequences of imaging, including radiation, incidental findings, and costs. We demonstrate that for patients with a risk of appendicitis <16% or >95%, imaging is not cost-effective. For patients at moderate risk >16%, an ultrasound without secondary signs of RLQ inflammation can be sufficient even if the appendix is not visualized and even in settings where the appendix is infrequently visualized. For patients with higher risk of appendicitis, providers should tailor their imaging approach for patients on the basis of a patient’s pretest probability of appendicitis.

We thank Wren Haaland, MPH, of Seattle Children’s Hospital, who created early iterations of the model and Matthew Dellinger, MD, and Joel Tieder, MD, MPH, of Seattle Children’s Hospital and the University of Washington School of Medicine, who reviewed the article.

CPT

Current Procedural Terminology

CT

computed tomography

ED

emergency department

ICER

incremental cost-effectiveness ratio

KID

Kids’ Inpatient Database

PedsQL

Pediatric Quality of Life Inventory

QALY

quality-adjusted life-year

RLQ

right-lower quadrant