Assessing adherence and cost-benefit of colorectal cancer screening for accountable providers

METHODS

In this retrospective electronic medical record review, the initial search screened for patients aged 50 to 75 years who completed FIT between June 1, 2014, and June 1, 2016 (Figure 1). Exclusion criteria were (1) not having an assigned primary care physician, (2) having a positive FIT test or colonoscopy within the year prior to completion of the FIT test results used for this study, and (3) dying within the follow-up period. Two authors independently screened patient charts and abstracted the following data: order date for FIT in outpatient setting, electronic documentation of communicating abnormal results to the patient, the ordering of and completion of diagnostic colonoscopy, the ordering of and completion of gastroenterology clinic visit, and the time to and results of diagnostic colonoscopy. Additionally, the following demographic and protected health information data were extracted: age, gender, ethnicity, insurance status, ZIP code, and previously positive FIT. This was an exploratory investigation, so a smaller sample was used as a proof of concept. Simple random sampling was conducted and used to explore the relation between abnormal FIT tests and rate of follow-up within 1 year. Measures of central tendencies were used to describe results.

A CBM explored the relation between monetary benefit of screening adherence via the (1) follow-up rate after abnormal FITs and (2) rate of colonoscopies after abnormal FITs. The primary outcome of interest was total monetary benefit in terms of avoided direct medical costs within 1 year of diagnosis, less the cost of screening, complications, and missed incidences. A population health perspective was chosen because health systems are increasingly taking accountability for a population, and assessing the monetary impact of various innovations on a population is an important tool for investment purposes. Additionally, the analysis focused on initial treatment costs instead of continuing, last year of life, or lifetime costs for two primary reasons: (1) annual FIT screening aligns with first-year costs and (2) payment model performance is typically calculated annually and therefore aligns with annual screening.

Financial calculations were carried out in a separate Excel file. Medicare’s 2017 physician, clinical diagnostic library, and anesthesia fee schedules were used with national payment instead of geographic-specific amounts (Table 1). The most frequent complications and rates associated with colonoscopy were estimated with diagnosis-related group codes using the Texas PricePoint tool (Table 1).^37,38 Authors identified six CRC treatment cost studies, inflated the average reported in the studies to 2017 dollars using the Consumer Price Index, and calculated the mean cost.^38,39 The cost of distant (stages 3 and 4) CRC ($42,039.70) treatment was adjusted downward, as demonstrated by Zhehui et al, to obtain the cost of local (stages 1 and 2) CRC treatment by 0.9204.⁴⁰ In constructing the CBM, multiple assumptions were made and are listed under corresponding tables. Further, dysplastic polyps were categorized as local CRC, adenomatous polyps were estimated to have a 25% annualized risk of progression to CRC, and hyperplastic polyps were estimated to have a negligible risk of progression to CRC.^41–47 Findings from the quality improvement/statistical analysis were used to inform the cost-benefit analysis.

A sensitivity analysis was conducted to investigate the robustness of benefit generated (Figure 2). In the first simulation, we allowed the rate of follow-ups to vary between 55% and 85%, holding all other variables constant. In the second simulation, we allowed both the rate of follow-up and the rate of follow-up colonoscopies to vary between 55% and 85% and 65% and 90%, respectively. In addition, the second simulation allowed the rate of complications to vary between the bounds of the confidence intervals of their incidence rates based on the literature.³⁸ In both simulations, the panel of patients was grown from 94 to 1000 patients to better fit rates of complications.

RESULTS

The initial search revealed 5675 FITs with results that were randomized and divided among the abstractors (Table 2, Figure 1). Of these, 1007 charts were screened for abnormal results, revealing 232 abnormal FIT results. The first 94 charts with an abnormal FIT test that were ordered for colonoscopy screening were included in this analysis.

In total, 64 of the 94 patients received appropriate follow-up care; however, less than half (45/94, 47.87%) of all patients received a follow-up diagnostic colonoscopy. Sixteen of the 94 patients (17.02%) received communication regarding abnormal results and were offered a colonoscopy but declined. Three of the 94 patients (3.19%) received communication regarding abnormal results, but health precluded them from being offered a colonoscopy. Other results are indicated in Table 2.

Approximately one-half of colonoscopies (24/45, 53.33%) had an abnormal colonoscopy finding, with 3/24 (12.5%) dysplastic polyps, 14/24 (58.33%) adenomatous polyps, and 7/24 (29.17%) hyperplastic polyps. Among the three patients with dysplastic polyps, one was eventually diagnosed as a stage 3 malignancy after a wait of 90 days from abnormal FIT to diagnostic colonoscopy. Another was diagnosed as a stage 4 malignancy after a 719-day wait from positive FIT to diagnostic colonoscopy. The third case of dysplastic polyp was attributed to hemorrhoids; when the patient presented with a complaint of anemia with unintentional weight loss approximately 2 years following the abnormal FIT, the diagnostic colonoscopy was performed.

The primary CBM demonstrated a population health benefit of $32,926 in the annual screening by avoiding possible CRC (Table 3). The largest cost components of CRC screenings were the average cost of a colonoscopy with any intervention ($9441) and anesthesia ($8928). The two patients diagnosed with distant CRC contributed an estimated $84,079.40 to costs. Total direct medical costs avoided by screening the 94-patient panel was $189,178, which was generated by 1 dysplastic plus 3.5 adenomatous (14 patients multiplied by 25% estimated rate of progression to malignancy as described in Methods section) for 4.5 total patient equivalents who received a colonoscopy after an abnormal FIT and had adenomatous polyps removed.

A secondary CBM was performed to ascertain the possible maximum benefits and costs for the studied panel of patients (Table 4). In this evaluation, each of the 91 patients received a colonoscopy (excluding an equal proportion of patients who would not be healthy enough for the procedure), and equal proportions from the improvement analysis were assumed to have abnormal colonoscopies. The costs associated with the expected number of patients with pre-existing local or distant CRC were added to the costs side of the model. Specifically, 53.33% of the 91 patients receiving a colonoscopy would have abnormal findings, meaning that 42 patients would have a colonoscopy without an intervention and 49 patients would receive a colonoscopy with an intervention. Further, whereas each of the 49 patients who received a colonoscopy with intervention would have a pathology, only 9.17 patients (2.09 dysplastic and 7.08 adenomatous patient equivalents) would receive the full cost avoidance benefit of screening. In this situation, total benefits through costs avoided increased from $189,178 to $385,363. Total annual benefits less costs were calculated to be over $63,237 for the year in which the patients were screened.

In a follow-up sensitivity analysis, the rate of follow-up colonoscopies appears to be the main driver in costs avoided/savings instead of overall rate of follow-ups (Figure 2). Total complication rates were not a major driver of cost; however, at the model’s minimum follow-up and colonoscopy rates, the overall costs appear greater than the benefits.

DISCUSSION

This two-part evaluation of CRC screening in an integrated health system demonstrates that improving screening guideline adherence can increase CRC cost avoided and, therefore, total monetary benefit for population. We found that though 68.09% of patients received appropriate follow-up care, in terms of being communicated results and offered a colonoscopy, only 47.87% of all patients had a follow-up colonoscopy as recommended by current treatment guidelines. For the assessed population, this equated to a benefit of over $189,178, less $33,478 in screening and $122,773 in two patients with distant and local CRC. Had CRC screening guidelines been followed appropriately, the net benefit would have exceeded $63,237. Additionally, a population simulation of 1000 patients that examined the sensitivity of monetary benefits across follow-up rates resulted in total monetary benefit, demonstrating that as follow-up and colonoscopy rates improved, overall monetary benefit significantly improved, because costs appear largely fixed or stagnant. Said differently, benefits of improving screening far outweigh the costs and demonstrate the need to incentivize higher rates of follow-up and colonoscopies to avoid costs.

Though follow-up rates fall within the range cited in the literature and above the national average for 2016 performance results for accountable care organizations (ACOs; ACO #19, 61.52%) and merit-based incentive payment system benchmark averages (ACO #113/PPRNET18, 57.48%), these results are underwhelming, especially given the increasing incidence of CRC in younger populations, the increasing number of baby boomers retiring into Medicare, and more pressure from payers and health systems to improve follow-up. A recent analysis of a large integrated urban safety net system in North Texas had similar results, where 42.30% of its study cohort failed to undergo a colonoscopy 1 year after positive screening compared to our 52.13%.²³ In that study, however, 57% of cases of failing to undergo a colonoscopy were attributable to patient failures, primarily patient refusal, compared to our 17% of patients who refused and 3% who were unable to undergo a colonoscopy because of poor health. The remaining 30% of cases that failed to receive appropriate follow-up colonoscopies were due to provider or system-level failures. This suggests an opportunity to improve adherence through improved outreach and patient and physician education. Addressing patient preferences, such as accuracy of the test versus frequency of testing, through shared patient-provider decision making may increase compliance.⁵⁰

To our knowledge, this is the first evaluation that utilizes a CBM from a population health perspective to explore the financial impact of CRC screening rates. Though there is a rich literature on patient-level benefits and cost-effectiveness studies of CRC screenings, no analysis has attempted to measure the potential monetary benefits to health system or production efficiency potential of appropriate screenings.^{13–15,38,51,52} Our results indicate that despite only 47% of patients receiving follow-up colonoscopies after abnormal FITs, the total benefit net of costs exceeded costs from complications or increased resources devoted to screening.

Several studies have already explored the relationship between new delivery models (ACOs, patient-centered medical homes, and integrated delivery systems) and cancer screenings.^54–57 However, more policies, pilots, and evaluations of the aforementioned are necessary to refine new care and payment models where cost-avoided dollars could be accounted for vis-à-vis shared savings. Though population health screening rate quality measures are already in use by payers and ACOs, policymakers should consider more closely linking CRC screening rates with alternative payment models.^48,49 One example would be using shared savings of avoided CRC costs as an incentive to improve compliance. This would involve shifting screening from a process measure to more of an outcome measure, because measures could be linked with incidence rates to ensure that incentive payments are appropriately rewarded. Through these alternative payment models, benefits could therefore be reallocated by ACOs and other community health providers, who take accountability in other areas like diabetes prevention, community interventions for asthma, and reductions in obesity rates.

A limitation to this study relates to external validity (generalizability), in that only charts within the system (limited by geography, hospital not-for-profit status) were selected. Results may be biased because random sampling was used instead of cluster sampling based on ordering physician; the authors did not investigate whether there was a relationship between follow-up rates and ordering physician. Moreover, the sample size was too small to perform formal statistical analysis. Patients who died were excluded due to the time frame of analysis. Additionally, complication rates were not collected, and estimates for the incidence rates and costs associated with each complication had to be estimated from previous literature using populations that may not exactly match the population in this study. We could not find data on what percentage of precancerous lesions, if not removed, would eventually be discovered as local CRC versus distant CRC. Thus, we calculated that all of these would be discovered as distant CRC, although this is probably not the case. Although we did our best to stay true to appropriate estimations for hospital costs, many of the primary source estimations vary widely. Our analysis did not consider the cost-benefit scenario of uninsured or underinsured populations. This population is less likely to receive CRC screening, and any malignancies found tend to be in a later stage (distant CRC) compared to malignancies found in those with health insurance. However, many hospitals, physicians, and pharmaceutical companies provide a degree of charity care annually, and it would not be unreasonable to think that providing charity care for screening would be more cost-effective from a population health and integrated health system perspective. Finally, this investigation assumed that all lesions were removed through endoscopy, but larger lesions may need to be removed through colectomy. However, this is relatively uncommon compared to endoscopic removal and, given the small size used in this investigation, would not be likely to impact the financial outcome.

Colorectal cancer is a devastating, prevalent, and expensive disease, but appropriate screening can significantly reduce its incidence and mortality. Adherence to screening guidelines, however, remains suboptimal. Though the assessed rate of follow-up after abnormal FIT (68.09%) is consistent with that found at other institutions, there is room for improvement, especially given the likely increase in new incidence and mortality rate. Health systems, ACOs, and payers should work together to structure financial incentives in a way that rewards providers for improved outreach and adherence guidelines. Optimized CRC screening could translate to the allocation of health care dollars toward other diseases.