Cost-Effectiveness of Community-to-Clinic Tailored Navigation for Colorectal Cancer Screening in an Underserved Population: Economic Evaluation Alongside a Group-Randomized Trial

Abstract

Purpose:

Although screening for colorectal cancer (CRC) lowers mortality and morbidity and is generally cost-effective, little is known about the cost-effectiveness of screening promotion.

Design:

Cost-effectiveness analysis alongside a group-randomized trial.

Setting:

Multicultural, underinsured communities in the Phoenix, Arizona, area.

Subjects:

English or Spanish speaking adults who were out of compliance for CRC screening guidelines.

Intervention:

All participants received community-based group education (GE) and the intervention group also received tailored community-to-clinic navigation (GE+TN).

Measures:

Number of participants screened and costs of tailored navigation, clinic visits, and CRC screening tests.

Analysis:

Incremental cost per additional person screened from the perspective of the healthcare system with bootstrapped confidence intervals.

Results:

Community sites were recruited and randomized to GE (n=120) and GE+TN (n=119). Across these sites 1,154 individuals were screened, 504 were eligible, and 345 attended the group education class (n=134 GE; n=211 GE+TN). Screening rates (26.5% GE+TN; 10.4% GE; 16.1% increase 95% CI: 7%, 23%) and costs per participant ($271 GE+TN; $167 GE;. a net cost increase of $104 95% CI: $1, $189) were significantly higher in the intervention group. Incremental cost-effectiveness was $646 (95% CI: -$68, $953) per additional person screened.

Conclusion:

Depending on the value placed on an additional person screened, the addition of community-to-clinic tailored navigation to a community-based CRC screening promotion program may be highly cost-effective.

Purpose

Despite the decline in colorectal cancer (CRC) incidence and mortality in the US over time, CRC remains the second-leading cause of death from cancer for both men and women.¹ Regular screening facilitates the early diagnosis of CRC and contributes to the reduction of CRC-related morbidity and mortality.² However, screening rates for the population in general are below national targets³; rates for minorities and low-income populations are even lower^{4, 5}; and those without a regular source of care have the lowest rates of all.⁴ Having received a referral for CRC screening is one of the strongest predictors of adherence,^6–10 but referrals are unlikely among those who have no clinical home, or rarely visit a primary care provider; factors that are more common among poor and minority populations.

It is generally accepted that CRC screening is cost-effective—that the cost of screening compared to the life-years saved is worthwhile.^{11, 12} However, studies that examine the cost-effectiveness of screening promotion programs,¹³ and especially of community-based programs, are rare.^{14, 15} Limited healthcare dollars are available to promote CRC screening, and these dollars should be spent on interventions and strategies that generate the highest screening rates for the resources expended, making it critical to evaluate costs for screening programs with or without patient navigation.¹⁶

This study answers the question: What is the cost per additional person screened of the addition of community-to-clinic navigation to a community-based education intervention to promote CRC screening in an underserved population in the Phoenix, Arizona, metropolitan area. We present a cost-effectiveness analysis based on a group-randomized trial of this intervention.

Methods

This community-based RCT included two navigation phases and targeted hard-to-reach, multicultural, underinsured communities.

Sample selection

Community sites were recruited and then randomized in blocks of 6–8 using the computer program Minimization-8, which uses a minimization method of assignment to control for potential confounders, in this case, age and ethnicity mix at each site.¹⁷ Participant recruitment at each site was through presentations, flyers, and word of mouth. Participants were eligible for the study if they were 50 years of age or older, English or Spanish speaking, and were out of compliance for CRC screening guidelines at that time,^{18, 19} either never screened or currently due (i.e., annual stool test, every 5 years for flexible sigmoidoscopy, or every 10 years for colonoscopy or accelerated schedule for high risk individuals).

All participants who were eligible and who completed the informed consent process including giving written consent received a 1-hour group education (GE) class that included information on CRC risk factors, dietary and physical activity guidelines for reducing CRC risk, and types of CRC screening. In Phase I, after the class, participants in the sites randomized to tailored navigation (GE+TN) were guided into primary care clinics to receive a healthcare provider referral for CRC screening. This community-to-clinic tailored navigation was delivered via phone calls with tailored messaging that addressed any barriers participants faced in making and keeping a clinic appointment. In Phase I, the control group only received group education (GE) and calls to determine clinic appointment status. Clinic appointments and clinic attendance were tracked in both groups and all those who kept a clinic appointment (no matter their randomized group) received tailored navigation to complete screening (Phase II). More details regarding the design of this trial are available in Clinical Trial Registration: NCT01853774; Unique Protocol ID: 5 R01 CA16239302. This study received approval from the Arizona State University (ASU) and the Ohio State University IRBs. The University of Arizona IRB ceded authority to the ASU IRB. Approval No. 1207008010.

Effectiveness

For the purpose of the cost-effectiveness analysis Eeffectiveness was measured in terms of the percent of participants in each randomized group who completed CRC screening. The target sample size of 440 participants was based on the team’s earlier studies²⁰ and power calculations by Hedeker²¹ to provide more than 85% power to detect a 50% difference in CRC screening completion. Once screening was reported, screening completion was verified through medical record review. More detail on the estimation of this trial’s effectiveness is found elsewhere.²²

Costs

Detailed information on the cost of the startup and implementation of this intervention are available in a separate report.²³ Because we want to determine the cost-effectiveness of the addition of community-to-clinic tailored navigation to an existing community-based intervention, this study focuses on the costs that differed between the randomized groups (GE+TN and GE). Because both groups received the group education class, costs per participant for each group were identical through that point. In this study we focus on the costs incurred by the health system and the organization providing the navigation after the education classes. Those costs include the cost of the Phase I community-to-clinic tailored navigation, the cost of the Phase II clinic-to-screening-completion navigation (offered to all those who had a clinic appointment), and the costs of clinic visits and screenings. Costs of providing navigation were captured from study time logs (labor costs) and expense reports (non-labor costs), and the costs of office visits and screening tests were provided by the participating clinics. Care was taken to exclude any costs specific to the research aspect of this study (e.g., time to obtain participant consent or to collect data for this analysis). All costs were adjusted to reflect 2014 US dollars (USD) using the non-seasonally-adjusted medical care Consumer Price Index.²⁴ Unit costs for staff time, office visits and screening tests are shown in Table 1.

Table 1.

Unit costs used in the cost-effectiveness analysis, 2014 USD

Cost type	Unit cost	Source
Community health navigator (per hour)	$22.83	Actual salary plus benefits for staff involved in the study
Community site liaison (per hour)	$20.26	Actual salary plus benefits for staff involved in the study
Project manager (per hour)	$46.63	Actual salary plus benefits for staff involved in the study
Clinic office visit (average per visit, range of quotes)	$28 ($23–$34)	Quotes given by study clinics
Stool tests (average per test, range of quotes)	$76 ($30–$106)	Quotes given by study clinics
Colonoscopy (average per test, range of quotes)	$1,483 ($1,057–$2,261)	Quotes given by study clinics

Analysis

This study analyzed the cost-effectiveness of adding Phase I community-to-clinic navigation to a CRC screening program from the perspective of the healthcare system and in terms of the cost per additional person screened. This perspective of analysis does not require the capture of costs from participating individuals. Costs include the labor and non-labor expense of providing the navigation, and the clinic visit and screening test costs for each group. Effectiveness was measured as the percent of participants screened. A bootstrap resampling technique (bias-corrected and accelerated bootstrap,^{25, 26} 1000 replications) was used to examine the uncertainty around the cost, effectiveness, and cost-effectiveness estimates. Based on this analysis we also generated a cost-effectiveness acceptability curve which shows the probability of the intervention being cost-effective given different values for an additional person screened. All calculations used Microsoft Excel 2010 (Redmond, WA).

Results

A total of 239 community sites were recruited and at these sites 1,154 individuals were screened for eligibility; 650 at the GE+TN sites and 504 at the GE sites. Figure 1 shows the flow of participants through the trial. Of the 345 participants (211 GE+TN and 134 GE) who attended the group education class, and thus, began Phase I, a total of 81 made and kept a clinic appointment (64 or 30.3% in the GE+TN group and 17 or 12.7% in the GE group) and 70 completed CRC screening (56 or 26.5% in the GE+TN group and 14 or 10.4% in the GE group).

Figure 1. Participant flow through the study.

Table 2 shows the characteristics of those in each group who attended the group education class. As can be seen, participants were predominantly female, White, Hispanic, and lower income. In general, the two groups were comparable. However, the control group was significantly older (M=62.0, SD=7.8 vs M=58.7, SD=7.2, t=4.06, p=.011, in the GE+TN group), more likely to have a regular provider (44.1% vs 56.0% in GE+TN, x²_(df-1)=4.64, p=.031), and more likely to have had cancer (12.8% vs 4.8% in GE+TN, x²_(df-1)=7.12, p=.012).

Table 2.

Participant characteristics by intervention group; all %(n) unless otherwise indicated

		GE+TN	GE Alone	Chi-square
		(n = 211)	(n = 134)	Testdf, p-value
Mean Age (SD)		58.7 (7.2)	62.0 (7.8)	t133 4.06, p<.001
Sex	Male	35.1% (74)	30.6% (41)
	Female	64.9% (137)	69.4% (93)	0.7381, p=.390
Race	African Am	11.0% (23)	9.0% (12)
	Amer Indian	4.3% (9)	0.7% (1)
	White	77.3% (163)	82.8% (111)
	Other	7.6% (16)	7.5% (10)	4.133, p=.248
Ethnicity	Latino/a	70.0% (147)	63.4% (85)
	Non-Latino/a	30.1% (63)	36.6% (49)	1.611, p=.204
Education	Less than HS	41.4% (87)	43.6% (62)
	HS/GED	21.8% (46)	15.7% (21)
	Some College	26.7% (56)	20.1% (27)
	Vocational	1.4% (3)	2.2% (3)
	4yr Bachelor	4.8% (10)	8.2% (11)
	Post Graduate	4.3% (9)	7.5% (10)	6.915, p=.227
Income	Less than 9,999	36.6% (59)	36.6% (37)
	10,000–24,999	36.4% (59)	36.6% (37)
	25,000–49,999	16.1% (26)	21.8% (22)
	50,000 +	11.2% (18)	5.0% (5)	3.823, p=.282
Relational Status	Partnered	49.3% (104)	41.0% (55)
	Had Partner	39.5% (83)	50.0 (67)
	Single	11.4% (24)	9.0% (12)	3.812, p=.149
Insurance	AHCCCS(Medicaid)	20.1% (27)	22.4% (47)
	Medicare	26.1% (35)	15.7% (33)
	Private Insurance	9.7% (13)	16.2% (34)
	None/Donť Know	44.2% (60)	45.7% (97)	7.283, p=.063
Regular Provider	Yes	44.1% (93)	56.0% (75)
	No	55.9% (118)	44.0% (59)	4.641, p=.031
Higher Risk	Yes	10.0% (21)	9.0% (12)
	No	90.0% (190)	91.0% (122)	0.7481, p=.387
Have/had cancer	Yes	4.8% (10)	12.7% (17)
	No	95.2% (201)	87.3% (117)	7.121, p=.018

The results of the cost-effectiveness analysis are found in Table 3. This table also shows the resource use underlying each cost component so that readers who face different unit costs than what are shown in Table 1 can easily adjust our results to their situation.²⁷ The total costs to the healthcare system (assuming the healthcare system pays for the delivery of this intervention—i.e., the navigation) per participant who attended the class were $271 for the GE+TN group and $167 for the GE group. The GE group’s costs are lower because they only received reminder calls and did not receive community-to-clinic navigation. They also had fewer clinic visits and screening tests than the GE+TN group. The increased cost per participant of the addition of community-to-clinic navigation was $104. This cost increase is compared to the increase in screenings (a net increase of 0.16 screenings per participant) to get a cost of $646 per additional person screened. The variation around this estimate is demonstrated by the confidence intervals shown in Table 3, but also by the cost-effectiveness acceptability curve (Figure 2).

Table 3.

Healthcare system resource use, costs and effectiveness per person by randomized group

	Resource use		Costs
	Tailored Navigation (GE+TN; n=211)	Control (GE; n=134)	Tailored Navigation (GE+TN; n=211)	Control (GE; n=134)	Incremental (TN+GE minus GE)
Phase I navigation (hours)	2.88	1.95	$65.87	$44.47	$21.41
Phase I clinic visits	0.33	0.13	$9.30	$3.56	$5.74
Phase II navigation (hours)	0.65	0.25	$14.92	$5.70	$9.21
Phase II stool tests	0.16	0.03	$11.87	$2.27	$9.61
Phase II colonoscopies	0.11	0.07	$168.69	$110.68	$58.01
TOTAL HEALTHCARE SYSTEM COSTS*			$270.65 ($212, $340)	$166.67 ($114, $257)	$103.98 ($1, $189)
Percent of participants screened*			26.5% (21%, 33%)	10.4% (7%, 18%)	16.1% (7%, 23%)
COST PER ADDITIONAL PERSON SCREENED*					$646.12 (−$68, $953)

^*95% confidence intervals calculated using bias-corrected and accelerated bootstrap.

Figure 2. Cost-effectiveness acceptability curve.

At a value of $500 per additional person screened the probability that the intervention would cost less than it is worth to the healthcare system is 23% under the cost assumptions used in this analysis. At a value of $750 per additional person screened the probability is 70% and at a value of $1,000 the probability is 97%.

To determine whether the cost of adding community-to-clinic navigation is worthwhile (i.e., whether the addition of community-to-clinic navigation is cost-effective), the value to the healthcare system of an additional person screened is found on the x-axis in Figure 2 and the height of the curve at that point indicates the probability that the cost of the intervention would be offset by its value (i.e., the probability that the intervention is net beneficial). At a value of $500 per additional person screened the probability is only 23% that the intervention would cost less than it’s worth to the healthcare system under the cost assumptions used in this analysis. At a value of $750 per additional person screened the probability is 70% and at a value of $1,000 the probability is 97%.

Discussion

This cost-effectiveness analysis estimated the additional costs and additional benefits that would incur to a community-based CRC screening program from the addition of community-to-clinic navigation. The incremental costs of this addition were estimated to be $104 per participant and the incremental benefits were an average of 0.16 more screenings per participant. The incremental cost effectiveness ratio is then $104/0.16 or $646 per additional person screened. An examination of sample variation using a resampling technique (bootstrapping) indicates that both the cost increase and the increase in screenings seen in the intervention (GE+TN) group were statistically significant, and the cost-effectiveness acceptability curve indicates a high (97%) probability that the cost per additional person screened would be worthwhile (net beneficial) as long as the value given to that additional screening was greater than $1,000.

Other studies have found that the addition of navigation results in net benefits: improved screening rates and lower overall healthcare costs than would be seen without the navigation. However, we found only one other study that examined the costs and effectiveness of using tailored community-to-screening navigation.^28–30 Similar to our intervention, this study targeted an underserved uninsured Hispanic population, and provided education and navigation to screening. Instead of having all participants make and keep a clinic visit, average-risk participants were given a stool test kit to use at home and a prepaid envelop with which to submit their kit for testing. High-risk participants were directed to the clinic to receive screening colonoscopy. The results of this program were compared to usual care (no organized community outreach) using state-transition Markov modeling of the results for a hypothetical cohort of 10,000 participants with characteristics similar to what was seen during the actual program.²⁸ Over a three-year time horizon, which included the costs of the program, the screening tests, and where appropriate, the first year of cancer treatment, the model estimated costs per participant for this tailored navigation program of $261 and for usual care of $86, for a net cost of $174 per participant in 2012/2013 USD. The model also demonstrated increases in screening rates by 23.4%. Because this was a budget impact analysis (looking out over three years rather than participants’ lifetimes) the study did not report an incremental cost-effectiveness ratio. However, a simple comparison of incremental costs ($174) to incremental benefits (0.234) gives a ratio of $744 per additional person screened.

Two other studies reported the cost-effectiveness of programs that started with a clinical population and navigated from clinic referral to screening. A study by Ladabaum et al³¹ used Markov modeling to show that clinic-to-colonoscopy screening navigation could improve screening rates by 15 percentage points (an average of 0.15 more screenings per participant) while lowering overall healthcare costs by $136 ($2558 - $2422) in 2012 USD compared to no navigation. This results in an incremental cost-effectiveness ratio of $907 per additional person screened. A second study used tailored navigation to help clinic patients who were due for CRC screening to get screened.³² The additional cost in 2011 USD for their tailored clinic-to-screening navigation intervention (excluding clinic visit and screening costs) compared to no navigation (usual care) was $289 per participant and the increase in screening rates was 24.7 percentage points (0.247 more screenings per participant) resulting in an incremental cost-effectiveness ratio of $1,172 per additional person screened compared to no navigation. Tailored clinic-to-screening navigation compared to their standard intervention (mailed booklet on CRC screening, a personalized letter with an information number, and stool test kit) increased costs by $122 per person and increased screening prevalence by 6.2 percentage points resulting in an incremental cost-effectiveness ratio of $1,958 per additional person screened. Depending on the value placed on an additional person screened, these clinic-to-screening navigation interventions could also be cost-effective.

Whether our currently reported navigation-from-community-to-clinic program is cost-beneficial depends on the value given to an additional person screened—a value which is difficult to estimate precisely. A rough estimate for this value can be made using estimates from other studies for the prevalence of an adenomatous polyp in a similar population (15%³¹) and for the difference in lifetime healthcare costs for those with screening-detected versus undetected polyps ($6,100 in total savings for males and $8,200 total savings for females in 2001 USD³³; approximately $9,700 and $13,100 in 2014 USD after adjusting using the medical care CPI²⁴). Multiplying the polyp prevalence by the cost savings gives an estimate of the value of an additional person screened of $1,460 for males and $1,963 for females. These estimates are both greater than $1,000; thus, indicating a high probability (>97%) that our addition of community-to-clinic navigation would generate net benefits for the healthcare system. These estimates are also likely conservative given the characteristics of the estimates used. Estimates of the prevalence of adenomas vary widely across studies,^34–37 but prevalence is well-known to increase with age. The estimate we used of 15% was for a 50-year-old adult and the average age of our screened population was 59 years.³¹ In addition, the estimated healthcare savings we used were for 65-year-old males and for 65-year-old females, and these costs are known to be higher for those younger.³³

In our analyses, we captured sample variation in office visit and screening rates using bootstrap resampling. We did not perform an analysis of the sensitivity of results to the cost parameters used because we lacked reasonable information on appropriate ranges for these costs—i.e., the ranges reported by our clinics were not representative of the ranges shown in other studies.²³ Instead we reported our results in enough detail so that the reader can easily determine the impact of varying any of these costs over any range. For example, over half of the cost increase when comparing the GE+TN group to the GE group was for the additional colonoscopies, therefore, the cost per additional person screened would be most sensitive to changes in the cost of colonoscopies. Our clinics reported costs of $1,057 to $2,261 per colonoscopy and we used an average cost of $1,483, but other studies have reported substantially lower costs.^38–40 The impact on the cost per screening of using some other cost per colonoscopy is determined by taking the incremental cost of the colonoscopies ($58, Table 3) dividing by the cost per colonoscopy used in this study ($1,483) and multiplying by the new cost per colonoscopy (say, $750) and then substituting this value ($29) into the sum to get a new cost per participant for the program of $75. This number divided by the increase in screenings (0.161) generates the new cost per additional person screened if the cost per colonoscopy was cut roughly in half of $468. Similar calculations will allow readers to determine the cost per additional person screened using unit costs appropriate for their setting.

While this study had many strengths, it also had limitations. Although every effort was made to properly exclude research-related costs, it is possible that costs that should not have been included may have inadvertently been included, or costs that should have been included may have been excluded. Although we captured sample variation through bootstrap resampling, we did not perform a formal sensitivity analysis of the cost parameters. We did, however, report our results in enough detail so that the reader can easily determine the impact of varying any of these costs over any range. This helps with the transferability of these results to other settings; the closest economic evaluations can get to generalizability.^{25, 27, 41, 42}

In summary, this article describes a cost-effectiveness analysis of the addition of community-to-clinic tailored navigation to a community-based CRC screening program targeting a hard-to-reach, multicultural and underinsured population. Community-to-clinic tailored navigation increased program costs, but also increased screening rates by 16 percentage points, and according to one estimate of the value of an additional person screened, provided net benefits to the health care system. Community-to-clinic navigation should be considered as an add-on to increase screening in other community-based CRC screening promotion programs.

So What? (Implications for Health Promotion Practitioners and Researchers).

• What is already known on this topic?

Colorectal cancer (CRC) screening is generally agreed to be cost-effective. However, screening rates are below national targets; rates for minorities and low-income populations are even lower; and those without a regular source of care have the lowest rates of all.

• What does this article add?

This study found that community-to-clinic tailored navigation significantly increased CRC screening rates in a low-income underserved population at a cost per additional person screened of $646. Depending on the value placed on an additional person screened, adding community-to-clinic navigation for CRC screening promotion may be highly cost-effective.

• What are the implications for health promotion practice or research?

Given the known benefits of CRC screening and the barriers faced by those with no clinical home, community-to-clinic tailored navigation should be considered as a cost-effective way to increase CRC screening rates.