Projections of demand and capacity for colonoscopy related to increasing rates of colorectal cancer screening in the United States
Dr S. Vijan, VA HSR&D, 2215 Fuller Road, Mail Stop 11H, Ann Arbor, MI 48105, USA.
Background : There is debate about the optimal colorectal cancer screening test, partly because of concerns about colonoscopy demand.
Aim : To quantify the demand for colonoscopy with different screening tests, and to estimate the ability of the United States health care system to meet demand.
Methods : We used a previously published Markov model and the United States census data to estimate colonoscopy demand. We then used an endoscopic database to compare current rates of screening-related colonoscopy with those projected by the model, and to estimate the number of endoscopists needed to meet colonoscopy demand.
Results : Annual demand for colonoscopy ranges from 2.21 to 7.96 million. Based on current practice patterns, demand exceeds current supply regardless of screening strategy. We estimate that an increase of at least 1360 gastroenterologists would be necessary to meet demand for colonoscopic screening undergone once at age 65, while colonoscopy every 10 years could require 32 700 more gastroenterologists. A system using dedicated endoscopists could meet demand with fewer endoscopists.
Conclusions : Colorectal cancer screening leads to demand for colonoscopy that outstrips supply. Systems to train dedicated screening endoscopists may be necessary in order to provide population-wide screening. The costs and feasibility of establishing this infrastructure should be studied further.
Colorectal cancer (CRC) is one of the leading causes of cancer death in industrialized nations,1 and screening to allow early detection and prevention of CRC can substantially reduce CRC morbidity and mortality.2–14 National organizations are in agreement that colorectal cancer screening is beneficial; guidelines for colon cancer screening in average-risk individuals recommend, starting at age 50 years, various combinations of faecal occult-blood testing (FOBT), flexible sigmoidoscopy (FS), barium enema, or colonoscopy.15–17 The common final pathway that provides the benefit of all of these screening tests is colonoscopy, which allows direct examination of the entire colon, biopsies of suspicious lesions, and removal of polyps that have malignant potential.
While recommendations for CRC screening have been in place for the better part of the last decade, there has been a recent increase in attention towards CRC screening. Celebrity spokespeople have testified before congress on the importance of screening and have undergone colonoscopy on live television.18 There are reports that there was a sharp increase in the demand for colorectal cancer screening immediately following these events.19 In many ways, the increase in publicity is a welcome and overdue phenomenon, as studies suggest that a disappointingly small percentage of the USA population currently undergo appropriate screening for CRC.20, 21
However, an increase in demand for CRC screening raises important questions about the ability of the health care system to meet this demand. In particular, the demand for colonoscopy, either as the initial screening test or as a follow-up for other positive tests (e.g. FOBT or FS), is of concern, as colonoscopy is usually performed by a limited pool of subspecialists. Indeed, there is now anecdotal evidence of long waiting lists for colonoscopy, and some facilities are limiting access to endoscopic screening because of limited supply.22 While there is little remaining doubt that CRC screening is likely to be cost-effective by current economic standards,7–14 few studies have quantified the costs or demand that will be placed on the health care system should a larger percentage of the USA population actually desire screening. Therefore, we sought to undertake an analysis of the number of screening procedures that would be necessary in the USA population equal to or over age 50, and analysed the ability of the health care system to meet the demand for these procedures.
We used a previously published Markov decision model to estimate the number of lifetime colonoscopies and flexible sigmoidoscopies underwent (per patient) under various screening strategies. The details of the model are reported elsewhere;14 briefly, the model simulates the natural history of colon cancer using polyp prevalence and incidence rates and cancer incidence rates. Screening strategies were tested, with efficacy based upon the accuracy of the tests that allow detection of cancer in early stages and prevention of cancer through removal of adenomatous polyps.
For this analysis, we used the decision model to estimate the number of colonoscopies and flexible sigmoidoscopies undergone using several screening strategies: flexible sigmoidoscopy combined with faecal occult-blood testing (FS/FOBT), a single colonoscopy at ages 50, 60, or 65 (an age which is later than most guidelines would accept but which we chose to model because Medicare covers screening colonoscopy, which will have direct impact on those who are uninsured prior to age 65), twice lifetime colonoscopy undergone at ages 50 and 60, or colonoscopy every 10 years. Positive results with screening FS/FOBT resulted in performance of colonoscopy. Patients in whom high-risk adenomatous polyps were diagnosed (polyps 1 cm or greater, three or more polyps, or polyps with advanced histology) underwent an initial surveillance colonoscopy at 3-years; if this endoscopy was negative, further surveillance was carried out at 5-year intervals. Lower risk polyps did not lead to surveillance; thus, our projections are conservative. We did not include either isolated FOBT or flexible sigmoidoscopy because our prior analyses suggested that these strategies were both less effective and more costly than colonoscopic screening strategies, in particular the once-lifetime strategies.14
The key assumptions for the model are outlined in Table 1. We used the current population of the United States, ages 50–79, as the cohort for the analysis, using the most recent census estimates to define the number of people in this age range.23 The cohort was divided into 5 year clusters and the model was run to assess the total number of colonoscopies and flexible sigmoidoscopies that would be necessary for each of these age clusters over the remaining lifetime of the cohort.
Table 1. Model assumptions
|US population (age)|
| 50–54||17 177 000||–|
| 55–59||13 284 000||–|
| 60–64||10 646 000||–|
| 65–69||9 430 000||–|
| 70–74||8 745 000||–|
| 75–79||7 414 000||–|
| 80–84||4 900 000||–|
| 85–89||2 696 000||–|
|Number of board-certified gastroenterologists in USA||11 044||10 000 – 12 000|
|Average number of colonoscopies per gastroenterologist per month||21||16–32|
|Current annual screening colonoscopies by gastroenterologists in USA||1 325 600||1 000 000 – 1 500 000|
|Proportion of screening colonoscopy done by non-gastroenterologists (%)||33||20–40|
We ran these calculations using two main scenarios. In the first, we assumed a fixed adherence level of 70%; i.e. 70% in the age-range were eligible for and followed recommendations for initial screening. While this is higher than current rates and is above even the goals set by the USA government,24 it is a reasonable target for screening in the long-term, based on rates of screening for other cancers.25 In the second, as an upper bound of demand estimates, we assumed that all patients were eligible, but that the total proportion of eligible and willing patients declined with age (a proxy measure for comorbidity). In order to define a range for this analysis, we analysed data from the Health and Retirement Study, a national household survey of USA adults aged 51–61 years in 1992.26 Although there is little information on the appropriate cutoff for screening eligibility, we assumed that patients with four or more chronic diseases had too many comorbidities to be reasonable candidates for screening. Using this assumption, we found that 92% of the population is screening-eligible at age 50, and that this number gradually declined to 64% by age 80. Age 79 was assumed to be the upper limit for screening eligibility. The demand for colonoscopy was estimated by using census estimates and the proportion of eligible people to generate predictions of the number of colonoscopies that would be required under the various screening strategies.
We also estimated the ability of the system to meet the demand predicted by the Markov decision model. To provide a basis for the number of screening procedures done currently, we used the Clinical Outcomes Research Initiative (CORI) database to determine rates and indications for colonoscopy among gastroenterologists in the USA. CORI is a national voluntary consortium of 400 gastrointestinal endoscopists at 42 sites in22 states.27 After a colonoscopy is completed, a computerized form is submitted electronically to a data repository in Portland, Oregon. Each colonoscopy report includes test date, a physician identifier, the indication for the procedure, and patient demographic information. This sample of gastroenterologists is not a true representative sample of the USA population of gastroenterologists, but does include a broad cross-section of practice types, ranging from academic to private practice, and provides a general description of current patterns of colonoscopy utilization.
We used the CORI data to calculate the average number of colonoscopies performed per month by USA gastroenterologists from October 2002 through September 2003. CORI is not a truly national representative database, and there may be biases in reporting (for example, those gastroenterologists who perform few or no procedures have little incentive to participate); however, at this time, it is the most complete ongoing accounting of endoscopy rates that is currently available.
We then examined the colonoscopies that were performed for colorectal cancer screening and surveillance based upon the indication reported in the CORI database. We defined colonoscopies that were done for primary screening (either in average risk individuals or those with a family history), follow-up of positive FOBT or flexible sigmoidoscopy, and surveillance for polypectomy or prior colorectal cancer as screening-related indications; each of these is counted in the Markov decision model.
The average rate of colonoscopy performance was then extrapolated to all USA gastroenterologists. At present, according to the American Board of Internal Medicine, there are 11 044 board-certified gastroenterologists in the USA.28 This allowed us to compute an approximate number of colonoscopies related to colorectal cancer screening that are performed in the USA each year. We then increased this number by 33%, based on the assumption that some screening-related colonoscopies that are performed by providers other than gastroenterologists, such as surgeons or family practitioners, as reported in the National Cancer Institute survey of colorectal cancer screening practices.29
We then calculated the incremental number of colonoscopies that would be performed under various screening scenarios by subtracting the number of screening-related colonoscopies that are done at present from the total number of colonoscopies predicted by the decision model. Subsequently, we examined the number of new specialists that would be necessary to meet this demand, first based upon a constant number of colonoscopies performed by the average provider, and then under various assumptions about the number of examinations that could be done on a monthly basis (for example, if there were dedicated technicians trained to do only colonoscopy and not take on other gastroenterological clinical duties). We also conducted sensitivity analyses varying the proportion of endoscopies undergone by specialties other than gastroenterology.
A consideration that may affect estimates of supply and demand is the possibility that an increase in colonoscopies due to screening may lead to a decrease in colonoscopies undergone for diagnostic purposes; i.e. patients who have had a recent colonoscopy may not be referred for another colonoscopy if they develop symptoms. There are no data to suggest whether or not this effect occurs, or the magnitude; we therefore examined the effects of such a shift by varying the proportion of current tests that are related to screening. In our base case, approximately 46% of colonoscopies are related to screening and/or surveillance of average risk individuals; we ranged this from 46 (e.g. no shift) to 66% (a 20% shift) in sensitivity analyses.
Current colonoscopy utilization
The average gastroenterologist in the CORI data set performs a mean of approximately 21 colonoscopies per month (range: 0–102), with 45.8% of these dedicated to screening (either primary screening, follow-up of an abnormal FOBT or flexible sigmoidoscopy, or for surveillance in patients with prior polyps or CRC). There is a very broad range of reporting in terms of monthly rates of colonoscopy, which presumably reflects the diversity of practice styles for gastroenterologists; however, as we were interested in total supply and demand issues, we used mean values rather than medians to calculate supply. Of note, the National Cancer Institute survey of colorectal cancer screening practices found that the average gastroenterologist performed about 32 colonoscopies per month; however, their sample was limited to those who performed procedures, and the proportion reporting screening-related colonoscopies in their sample was lower (39% vs. CORI estimates of 45.8%), although this may in part be due to reporting bias in their sample.
When extrapolated to the general population of gastroenterologists in the USA, this suggests that approximately 1.27 million colonoscopies are performed by gastroenterologists annually for colorectal cancer screening-related indications. Increasing this number by 33% to include colonoscopies not done by gastroenterologists leads to an estimate of 1.69 million screening-related colonoscopies each year. Although our data sources are different, these numbers compare well with those projected from the National Cancer Institutes survey of colorectal cancer screening practices, which estimated that there are about 1.6 million screening examinations in the USA on an annual basis.
Predicted colonoscopy demand
We then used the decision model to examine the potential demand for colonoscopy given different national CRC screening policies. Table 2 demonstrates the demand under our two main scenarios: a ‘reasonably’ achievable demand (70% screening adherence) and a ‘maximal demand’ strategy based on age-related eligibility (100% base adherence, less age-related comorbidity). In the 70% adherence scenario, the average patient will, in their lifetime, have between 0.92 (for a strategy of screening once with colonoscopy at age 65) and 2.43 colonoscopies (for a strategy of screening with colonoscopy every 10 years). In the maximal demand strategy, this estimate ranged from 1.04 to 2.84 colonoscopies per subject. Notably, the demand for colonoscopy with the FS/FOBT strategy falls in the mid-range of the demand with various colonoscopic strategies. Indeed, one of the major concerns about colonoscopic screening, high demand for colonoscopy, is only partly ameliorated by choosing FS/FOBT as a strategy; focused colonoscopic strategies can lead to fewer colonoscopies than FS/FOBT. Screening with FS/FOBT also leads to an average total of about four endoscopic procedures (including 2.3–3.7 sigmoidoscopies, depending on adherence) per patient over a lifetime.
Table 2. Estimated number of colonoscopies with each screening strategy
|Annual faecal occult- blood testing and flexible sigmoidoscopy every 5 years†||1.08||2.89||1.20||9.03||1.50||4.08||2.39||13.67|
| At age 50||1.11||3.53||1.84||3.53||1.41||4.51||2.82||4.51|
| At age 60||1.06||2.58||0.89||2.58||1.20||2.92||1.23||2.92|
| At age 65||0.92||2.21||0.52||2.21||1.04||2.49||0.80||2.49|
| At ages 50 and 60||1.94||5.51||3.82||5.51||2.33||6.70||5.01||6.70|
| Every 10 years||2.43||6.72||5.03||6.72||2.84||7.96||6.27||7.96|
Table 2 also presents the total number of colonoscopies that would be done on an annual basis, assuming a relatively steady-state population. Given the current estimate that 1.6–1.7 million colonoscopies are currently undergone for screening purposes in the USA annually, screening the entire age-eligible USA population would lead to an estimated demand for screening-related colonoscopy that is several times higher than it is at present. This fits fairly well with current estimates that between 21 and 34% of the USA population was screened appropriately during a recent national assessment.30
Table 3 demonstrates the incremental number of endoscopists that would be necessary to meet this demand, based upon various estimates of the number of colonoscopies that could be done monthly (at present, CORI data suggest that gastroenterologists perform an average of 21 colonoscopies per month; however, we view that as a low-end estimate as training more gastroenterologists would, in theory, reduce the clinical burden of other duties and free up time to do more colonoscopies). Even under the most favourable sensitivity analysis assumptions, a single colonoscopic screening at age 65, which requires the fewest number of colonoscopies, would require between 1360 and 4160 more gastroenterologists. The more effective but also more resource intensive strategies would require a three to four times increase from the current number (11 044) of gastroenterologists in order to meet demand. Again, however, even the FOBT/FS-based strategy would require a substantial increase in the number of gastroenterologists, ranging from a 30 to 110% increase, depending on assumptions about eligibility and the number of colonoscopies that could be performed.
Table 3. Incremental number of gastroenterologists needed to meet demand for screening-related colonoscopy*
|Annual faecal occult- blood testing and flexible sigmoidoscopy every 5 years||6270 (5810–7130)||12 440 (11 980–13 300)||4180 (3870–4750)||8290 (7990–8870)||3140 (2900–3560)||6220 (5990–6650)|
| At age 50||9610 (9140–10 470)||14 690 (14 230–15 550)||6400 (6100–6980)||9790 (9480–10 370)||4800 (4570–5230)||7340 (7110–7770)|
| At age 60||4670 (4200–5520)||6400 (5940–7260)||3110 (2800–3680)||4270 (3960–4840)||2330 (2100–2760)||3200 (2970–3630)|
| At age 65||2710 (2250–3570)||4160 (3700–5020)||1810 (1500–2380)||2780 (2470–3350)||1360 (1120–1780)||2080 (1850–2510)|
| At ages 50 and 60||19 930 (19 470–20 790)||26 120 (25 650–26 970)||13 330 (12 980–13 860)||17 410 (17 100–17 980)||9970 (9730–10 390)||13 060 (12 830–13 490)|
| Every 10 years||26 210 (25 750–27 070)||32 700 (32 230–33 550)||17 470 (17 170–18 050)||21 800 (21 490–22 370)||13 110 (12 870–13 540)||16 350 (16 120–16 780]|
Alternatively, if a group of dedicated endoscopists were trained (e.g. ones that had no clinical duties beyond endoscopy), a much smaller number of new endoscopists would be required to meet the projected demand (Table 4). For example, if colonoscopy every 10 years is the predominant strategy, and a dedicated endoscopist can perform 200 colonoscopies per month (about 10 per day), then between 2100 and 2600 endoscopists could theoretically meet demand. This assumes, however, that these endoscopists could be ideally geographically situated to meet demand; it is more likely that a somewhat higher number would need to be trained in order to ensure equity in geographic availability of procedures.
Table 4. Number of dedicated endoscopists needed to meet demand for screening-related colonoscopy*
|Annual faecal occult- blood testing and flexible sigmoidoscopy every 5 years||670 (620–760)||1330 (1280–1420)||500 (470–570)||1000 (960–1060)||400 (370–460)||780 (770–850]|
| At age 50||1030 (980–1120)||1570 (1520–1660)||770 (730–840)||1180 (1140–1240)||620 (590–670)||940 (910–1000]|
| At age 60||500 (450–590)||680 (630–770)||373 (340–440)||510 (480–580)||300 (270–350)||410 (380–460]|
| At age 65||290 (240–380)||440 (400–540)||220 (180–290)||330 (300–400)||170 (140–230)||270 (240–320]|
| At ages 50 and 60||2130 (2080–2220)||2790 (2740–2880)||1590 (1560–1660)||2090 (2050–2160)||1280 (1250–1330)||1670 (1640–1730]|
| Every 10 years||2800 (2750–2890)||3490 (3440–3580)||2100 (2060–2170)||2620 (2580–2680)||1680 (1650–1730)||2090 (2060–2150]|
We also examined the effect of a shift in colonoscopy utilization, with screening colonoscopies replacing those done for diagnostic purposes, on colonoscopy supply and demand. In the base case, we found that the system currently performs about 1.69 million screening-related colonoscopies annually. Shifting 10% of colonoscopies from diagnostic to screening leads to an estimate that the system can currently do 2.06 million colonoscopies for screening; shifting 20% increases this to 2.43 million. The effects of this shift on the incremental number of colonoscopies needed to meet demand is shown in Table 5. For the more aggressive strategies (e.g. colonoscopy every 10 years), there is a relatively smaller impact of this increase in supply, as the incremental number of colonoscopies is still substantial. In contrast, for less colonoscopy intensive programmes, a shift in the way we use colonoscopic resources substantially attenuate the increased demand; indeed, for once-lifetime colonoscopy at age 65, the system can already meet demand if there is a 20% shift from diagnostic to screening colonoscopy.
Table 5. Effects of shift from diagnostic to screening colonoscopy
|Annual faecal occult- blood testing and flexible sigmoidoscopy every 5 years*||2.89||1.20||0.83||0.46|
| At age 50||3.53||1.84||1.47||1.10|
| At age 60||2.58||0.89||0.52||0.15|
| At age 65||2.21||0.52||0.15||−0.22|
| At ages 50 and 60||5.51||3.82||3.45||3.08|
| Every 10 years||6.72||5.03||4.66||4.29|
Screening for colorectal cancer has been shown to be an effective and cost-effective intervention.9, 10, 12–14, 16 Unfortunately, adherence to screening recommendations remains disappointingly low.30 Attempts to increase adherence are important given the demonstrated effectiveness of colon cancer screening, but will inevitably lead to increased need for colonoscopy, regardless of the initial screening procedure; there is some evidence that this is already occurring.19, 22 Although colonoscopy is an increasingly accepted option for a primary screening test (including coverage by Medicare), the health care system appears to be unprepared to meet a demand for screening the full population of eligible people in the United States unless we adopt a new approach to selecting, training, and certifying endoscopists.
Indeed, our analyses clearly show that, even under optimistic assumptions, the system as currently structured cannot meet the projected demand for colonoscopy. One prior analysis of demand used somewhat different methods and assumptions but also found that projected demand outstripped current supply.29 Our analyses extend the findings by examining multiple different screening strategies under different adherence scenarios, as well as by conducting a basic examination of how the demand for colorectal cancer screening can be met. For example, a one-time colonoscopy strategy leads to considerably fewer colonoscopies than a combined FOBT/flexible sigmoidoscopy strategy. This may surprise some given that the latter strategy is often promoted as a way to ‘off-load’ a portion of screening to primary-care providers. Indeed, although this is the case, the number of sigmoidoscopies is also very large, and the total endoscopy rate is substantially higher with the FOBT/flexible sigmoidoscopy strategy than it is with any of our modelled colonoscopic strategies. Although twice lifetime colonoscopy, which our prior analyses have suggested is the most cost-effective strategy,14 leads to a substantial incremental demand for colonoscopy over the other strategies, our analyses suggest that a new policy strategy for increasing the supply of endoscopists is needed regardless of our primary approach to CRC screening.
However, our analyses also suggest that any shifting in resources from diagnostic to screening colonoscopy can help to attenuate some of this demand; the attenuation is most significant for once lifetime colonoscopy, particularly that done after age 60, and for combined sigmoidoscopy and faecal occult-blood testing. Whether this type of shift will occur, and the magnitude of such a shift, are currently unknown. It would require substantial efforts to reduce diagnostic testing in those who have had a previously negative colonoscopy, but if successful could defray some of the effort needed to ramp up colonoscopy supply. We would note, however, that the increasingly recommended primary strategy of screening with colonoscopy every 10 years is only somewhat affected even with substantial resource shifting; for example, even if we can reduce the number of diagnostic colonoscopies by an absolute 20%, and shift all of these resources to screening, we would still need to find capacity to conduct 4.3 million additional colonoscopies on an annual basis.
The up-front costs of such an approach are likely to be substantial. There will be initial training requirements as well as ongoing costs for salary, facility upkeep, and to replenish the ranks of practitioners who are dedicated to colon cancer screening. Further, more procedure units and facilities would need to be constructed to house these additional personnel. Our analyses suggest that there would need to be substantial increases in the number of gastroenterologists to meet demand for the most aggressive and effective screening strategies; given that there is likely not enough demand for other clinical duties for this many additional gastroenterologists, this seems a suboptimal approach. A more efficient strategy would be to train dedicated endoscopists. For example, advanced practice nurses can perform flexible sigmoidoscopy with accuracy, safety and patient satisfaction that is similar to that done by specialist physicians.31, 32 This type of strategy could theoretically be applied to colonoscopy as well, although the challenges would be greater. One could envision ‘screening centres’ where multiple endoscopists are supervised by an expert gastroenterologist, similar to a dialysis facility or a suite of operating rooms where nurse–anaesthetists are supervised and supported by physician specialists. Whether these endoscopists should be specialist physicians, generalist physicians, physicians’ assistants, nurse practitioners, nurse specialists, or technologists is an open question. Alternatively, demand could be met by increasing the average number of procedures performed by endoscopists. However, our analyses show that even increasing the average to 32 procedures per month is not adequate to meet projected demand; thus, we expect that unless substantial efficiencies can be achieved, the pool of endoscopists will need to be expanded.
There are several limitations to our analysis. We did not have true nationally representative figures for current procedure rates; instead, we relied upon estimates from CORI,27 a voluntary sample of gastroenterologists that may or may not accurately reflect current practice, and extrapolated these results to the broader population of gastroenterologists. However, we conducted broad sensitivity analyses around the number of procedures that could be performed in order to explore the possible effects of inaccuracy in these estimates, and all analyses supported the conclusion that large number of new endoscopists will be needed if we are to successfully achieve national CRC screening using currently recommended approaches. We also did not attempt to estimate the costs of training different personnel to meet projected colonoscopy demand. Finally, emerging technologies such as immunochemical FOBT, virtual colonoscopy, and screening with stool DNA markers may alter the demand equation as they are more widely tested and disseminated if proven effective.
The up-front costs of training more endoscopists and constructing facilities to house them are likely to be substantial, as are the ongoing costs of salary and facility upkeep. Although a number of prior analyses that suggest that screening for colorectal cancer is cost-effective, no prior analysis has included these costs; certainly at least a portion are likely to be included in current colonoscopy reimbursement rates.7–14 Because some screening strategies, particularly a single screening with colonoscopy are extremely cost-effective or even cost-saving under some assumptions, they are likely to stand up to more detailed analyses including the incremental costs of training and maintaining new personnel to meet screening demands. Regardless of how these changes in costs may affect the cost-effectiveness of CRC screening, however, it is clear that the health system cannot currently meet demands for screening-related colonoscopy, and that a new model for endoscopists is needed.
Dr Vijan is a VA Health Services Research & Development Career Development Awardee.