Estimating the preferences and willingness-to-pay for colorectal cancer screening: an opportunity to incorporate the perspective of population at risk into policy development in Thailand

Abstract

Aims

Colorectal cancer (CRC) is one of the public health burdens that can be lowered by early detection. This study aims to examine the preferences and willingness-to-pay of a population at risk for CRC screening in Thailand. Understanding the preferences for these individuals at risk would help Thailand, as an example of LMICs, to design effective population-based CRC screening programs.

Materials and methods

A discrete choice experiment (DCE) was conducted among screening-naïve adults aged 50–75 years, who were at risk of CRC, in the out-patient department of a tertiary care hospital in Thailand. A DCE questionnaire was developed from six CRC screening attributes. Each questionnaire was composed of six choice sets and each contained two alternatives described by the different levels of attributes and an opt-out alternative. Participants were asked to choose one alternative from each choice set. A multinomial logit model was developed to determine the relative preference of each attribute. The willingness-to-pays for all attributes and screening modalities and the estimated preferred choices of the annual fecal immunochemical test (FIT), 10-yearly colonoscopy, 5-yearly double-contrast barium enema (DCBE), 5-yearly computed tomographic colonography (CTC), 5-yearly flexible sigmoidoscopy (FS), and no screening was calculated and compared.

Results

Four hundred participants were included. All attributes, except pain and less bowel preparation, were statistically associated with the participants’ preference (p < .05). They preferred screenings with a high-risk reduction of CRC-related mortality, no complication, 5-year interval, and lower cost. The estimated preferred choices of FIT, colonoscopy, DCBE, CTC, and FS were 38.2%, 11.4%, 14.6%, 9.2%, and 11.4%, respectively. The willingness-to-pays for each screening modality was US$251, US$189, US$183, US$154, and US$142 (8,107, 6,105, 5,911, 4,974, and 4,587 THB) per episode, respectively.

Conclusions

The risk reduction of CRC-related mortality, complication, screening interval, and cost influenced the CRC screening preferences of Thai adults. FIT was the most preferred. Policymakers can develop a successful CRC screening campaign using these findings, incorporating the perspective of the population at risk in policy formulation to accomplish their goals.

Keywords:

• Preferences
• willingness-to-pay
• colorectal cancer screening
• discrete choice experiment
• Thailand

Introduction

Health technology assessment (HTA) is recognized as a crucial discipline for evidence-informed policy decision making under universal health coverage¹. It has been widely used for health benefits package design in many countries². However, HTA fails to capture various important value elements, especially the preferences of individuals^3,4. The preference is a key factor to determine the attendance and the success rate of policy implementation. Incorporating these individuals’ preferences into the decision-making process would better capture the value of health technology. Despite the recognition of the importance of this perspective, the practice of incorporating individual preferences in HTA remains uncommon, especially in low- and middle-income countries (LMICs). One major reason is that there is a lack of research and evidence on preferences^5,6.

Colorectal cancer (CRC) is one of the leading causes of cancer-related death globally^7–12. This public health burden can be lowered by screening policy⁹. The US Preventive Services Task Force 2016 recommended CRC screening among population aged between 50–75 years, which refer to “the population at risk,”¹¹ but recent evidence suggested that less than one-tenth of the global population adhered to this recommendation^13–18. The challenges in the policy implementation were related to the perceived effectiveness, costs, burdensome procedures, and their fear of pain and associated complications^19–25. As CRC is also recognized as a leading cause of death in the Thai population, policymakers and clinicians have a vast interest in determining a suitable CRC screening program for nation-wide implementation²⁶. Currently, CRC population-based screening campaigns using either fecal immunochemical test (FIT) or colonoscopy have been launched to assess implementation feasibility in Thailand. However, the screening participation rate of FIT and colonoscopy were only 63% and 47%, respectively^18,27. A recent study was conducted to assess the cost-effectiveness of CRC screenings in Thailand and found that colonoscopy provided the best value for money while the FIT was more affordable²⁸. However, the preferences of the individuals at risk of CRC were not included and their willingness-to-pay (WTP) was never examined. The information about their preferences and WTP would not only fill the gap of QALY-based value assessment in HTA but also help policymakers to increase the acceptance of the CRC screenings that were relatively low in the country²⁷. This study, therefore, aims to examine the preferences and their WTP of individuals at risk of CRC in Thailand by using a discrete choice experiment (DCE). Understanding the preferences for these individuals at risk would help Thailand as an example of LMICs to design effective population-based CRC screening programs.

Methods

Attribute and level identification

Two previously published reviews suggested that the number of attributes ranged between four to seven in most DCE studies^29,30. To select CRC screening attributes and levels, which were important in the decision to participate in a CRC screening program, we conducted a literature review based on the background of hypotheses from clinical knowledge. A total of 10 attributes were identified. Interviews with gastroenterologists were conducted to confirm these attributes. Then, we purposively selected 20 individuals at risk for in-depth interviews to confirm the attributes that were important to them. In addition, they were asked to rank the importance of these attributes. They consistently agreed that pain, risk reduction of CRC-related mortality, risk of complications, screening interval, bowel preparation, and out-of-pocket cost were important to them. Therefore, these six attributes were included in this study. Literature suggested that DCE must sufficiently vary the attribute levels^31,32. This study, therefore, identified the extreme ranges of the attribute levels from the literature review of all existing CRC screenings, as shown in Table 1. These levels were equally spaced. Table 2 summarizes the selected attributes and their levels.

Table 1. Screening programs’ characteristics.

Choice	Colonoscopy	Flexible sigmoidoscopy	DCBE	CT colonography	FIT	No screening
Pain	Mild pain	Mild pain	Mild pain	Mild pain	No pain	No pain
Risk reduction of CRC-related mortality^33	50–70%	22–42%	40–50%	40–50%	40–70%	0%
	(61%)	(33%)	(45%)	(45%)	(59%)
Risk of complications	Small	Small	Small	Small	No	No
Screening interval	10 years	5 years	5 years	5 years	1 year	0
Bowel preparation	Fluid 2 liters & 6-hour fasting	Laxative and enema	Laxative and enema	Fluid 2 liters & 6-hour fasting	No	No
Out-of-pocket cost (US$)	108–186	62–139	71–124	124–186	0.7–3	0
(THB)	3,500–6,000	2,000–4,500	2,300–4,000	4,000–6,000	25–100	0

Abbreviations. CRC, colorectal cancer; DCBE, double-contrast barium enema; FIT, fecal immunochemical test.

Table 2. Attributes and levels for colorectal cancer screening.

Attributes	Levels
Pain	No, mild
Risk reduction of CRC-related mortality (%)	0, 40, 80
Risk of complications	No, small
Screening interval (years)	1, 5, 10
Bowel preparation	No, minimal, intensive
Out-of-pocket cost (US$)	0, 93, 186
(THB)	0, 100, 6,000

Discrete choice experiment questionnaire development

It was not feasible to present all possible 432 (2² × 3³ x 4¹) combinations of the selected attributes and levels to study participants. An orthogonal and level balance design was used to randomly draw a fraction of all combinations by using Ngene 1.X softwareⁱ. A total of 36 choice sets were generated and randomly divided into six blocks. Each block comprised of six choice sets that were included in a questionnaire. Therefore, this study had a total of six different questionnaire versions. Each choice set contained three unlabeled alternatives, including two hypothetical CRC screening methods and an opt-out alternative. The opt-out alternative was used to resemble a real-world option since individuals might not choose any CRC screening test at all. The choice set for validity check was the choice set that composed of 3 unlabeled alternatives, one of which was the alternative that obviously worse than another one that contained all attributes with favorable levels (a dominant alternative) including lowest pain, highest risk reduction of CRC-related mortality, lowest complications, highest interval, and lowest out-of-pocket cost, and the other was an opt-out alternative. Those participants who understood the questionnaire would either choose the alternative that was obviously better than the others or the opt-out alternative.

In addition, questions on participant’s characteristics and experiences related to CRC and CRC screening were included in the questionnaire (Supplementary Table 1E). A think-aloud method was conducted with five individuals at risk of CRC to examine the understanding of the questionnaire (Supplementary Table 4E). Three medical doctors were asked to check the content validity of the questionnaire. In addition, we also performed a pilot test on a convenient sample of 30 population at risk individuals for the questionnaire (Supplementary Table 3E). No major change was made.

Data collection

The study protocol was approved by Siriraj Institutional Review Board No. 298/2560 (EC1). This study used multiple approaches, including a good DCE research practice and a published practice guide for achieving the statistical power of 80% to determine the study sample size^29,34. There is no optimal method for determining the sample size for DCE. The sample sizes of previous DCE studies generally ranged from 150 to 1,200 participants. A literature review previously showed that the underlying symptoms of individuals affected their preference for cancer screening³⁵. Therefore, we planned to show a comparison between the preferences for CRC screening among the individuals with gastrointestinal (GI) symptoms (such as abdominal pain, lower GI bleeding, bloating, constipation, diarrhea, and bowel habit change) and the preferences of the individuals without these GI symptoms. To accommodate the comparison of the preferences of these two groups, a literature review suggested a sample size of 200 per group³⁶. Therefore, a total of 400 screening-naïve adults, who were population at risk, including 200 individuals with GI symptoms and 200 individuals without GI symptoms, aged between 50–75 years who visited the out-patient department of the Faculty of Medicine Siriraj Hospital (a 2,061-bed hospital), Mahidol University³⁷ were purposively selected. We also oversampled for 10 percent (40 individuals) to ensure enough samples if any possible incomplete and invalid data existed. We excluded those participants who failed to correctly respond to the validity choice set. They were informed about the overall study, general background information of CRC and CRC screening, cost of each screening method and were then invited to participate in the study. One of the study investigators, a trained research assistant, used the developed questionnaires with a fact sheet to individually face-to-face interview the participants from October 2017 to January 2018. Each participant was asked to respond to one version of the questionnaire that consisted of three parts. The first part was to collect the demographic data. The second part included the questions that asked for their preferred choices of screening under two conditions, that is, free of charge and out-of-pocket payment with the willingness-to-pay amount for each method. Finally, they were asked to respond to one of the six DCE questionnaire versions, including the choice set for the validity test.

Data analysis

Only data from participants who correctly chose the right alternative in the validity choice set were included in the analyses. Participants’ characteristics and experiences in CRC were descriptively analyzed. Based on Random Utility Theory, participant responses for each choice set were observed and analyzed in DCE³⁸. The following utility, that a participant i assigns to an alternative j in a choice set s, U_isj, was estimated: $$U_{\mathit{isj}}={\beta }_0+{\beta }_1{\text{Pain}}_{\mathit{isj}}+{\beta }_2{\text{Mortality}\mathrm{ }\text{reduction}}_{\mathit{isj}}+{\beta }_3{\text{Complications}}_{\mathit{isj}}+{\beta }_4\text{Interval}{1}_{\mathit{isj}}+{\beta }_5\text{Interval}{5}_{\mathit{isj}}+{\beta }_6\text{Interval}10_{\mathit{isj}}+{\beta }_7\text{Preparation}{1}_{\mathit{isj}}+{\beta }_8\text{Preparation}{2}_{\mathit{isj}}+{\beta }_9{\text{Cost}}_{\mathit{isj}}+{\epsilon }_{\mathit{isj}}$$ where β₀ is the constant reflecting participant preference for CRC screening relative to no screening, β₁, β₂, β₃, β₄, β₅, β₆, β₇, β₈, β₉ are the coefficients or the mean attribute weights of pain, % risk reduction of CRC-related mortality (mortality reduction), complications, interval 1 year (interval 1), interval 5 years (interval 5), interval 10 years (interval 10), minimal bowel preparation (preparation 1), intensive bowel preparation (preparation 2), and out-of-pocket cost (cost), respectively, ε_isj is error term.

The multinomial logit model (MNL) was used to develop the utility model by using Nlogit 6ⁱⁱ. The level of statistical significance was set at 0.05. Marginal WTPs of the attributes were calculated by taking the ratio of the mean attribute coefficient to the mean coefficient of the cost attribute. Krinsky and Robb’s method was used to estimate 95% confidence intervals of WTPs of the attributes³⁹. Finally, WTPs for all CRC screening methods in the real-world practice was calculated by multiplying the marginal WTP for each attribute with the difference between attribute levels, which were obtained from clinical literature. All costs were converted and reported in United States Dollars (US$) using the averaged conversion rate in 2018 (1 US$ = 32.3 Thai Baht)^39,40. In addition, the preferences and WTP of the participants with and without GI symptoms were calculated.

As the goal of this study is to understand the preferences on the acceptability of the individuals at risk for the CRC screening methods, we calculated the percentage of estimated preferred choices of each screening test by using the following formula: $$\text{Probability\hspace{0.17em}of\hspace{0.17em}screening\hspace{0.17em}test\hspace{0.17em}A\hspace{0.17em}}=\frac{e^{\left[\right(\beta \mathrm{pain}\times \text{\hspace{0.17em}pain\hspace{0.17em}of\hspace{0.17em}test\hspace{0.17em}A})+(\beta \mathrm{cost}\times \text{\hspace{0.17em}cost\hspace{0.17em}of\hspace{0.17em}test\hspace{0.17em}A})+(\beta \text{risk\hspace{0.17em}reduction\hspace{0.17em}}\times \text{\hspace{0.17em}risk\hspace{0.17em}reduction\hspace{0.17em}of\hspace{0.17em}CRC\hspace{0.17em}mortality\hspace{0.17em}of\hspace{0.17em}test\hspace{0.17em}A})+(\beta \mathrm{complication}\times \text{\hspace{0.17em}risk\hspace{0.17em}of\hspace{0.17em}complication\hspace{0.17em}of\hspace{0.17em}test\hspace{0.17em}A})+(\beta \mathrm{interval}\times \text{\hspace{0.17em}interval\hspace{0.17em}of\hspace{0.17em}test\hspace{0.17em}A})+(\beta \mathrm{preparation}\times \text{\hspace{0.17em}preparation\hspace{0.17em}of\hspace{0.17em}test\hspace{0.17em}A}\left)\right]}}{\text{Summation\hspace{0.17em}of\hspace{0.17em}exponential\hspace{0.17em}of\hspace{0.17em}every\hspace{0.17em}test\hspace{0.17em}and\hspace{0.17em}no\hspace{0.17em}screening}}$$

Results

Out of 484 invited, 440 (91%) agreed to participate in our study. A total of 428 participants (97.3%) completed the questionnaire. Of these, 28 participants (6.5%) failed to correctly respond to the validity choice set, leaving 400 participants to be included in the study analyses.

Table 3 shows the characteristics of participants. The average age of participants was 62.4 ± 6.4 years. Two hundred and forty-six (61.5%) were female. The common comorbidities were hypertension, dyslipidemia, diabetes mellitus, and cardiovascular disease (51.0%, 31.0%, 24.0%, and 6.0%, respectively). More than half of the participants were retirees. The average monthly income was US$495 (US$217–929) (16,000 THB [7,000–30,000 THB]). More than half (53.5%) of the participants were aware of CRC. Approximately, one-fourth (23.5%) of the participants were aware of CRC screening. One hundred and sixty-two (40.5%) participants were not apprehensive about CRC. When the participants were directly asked, 47.0%, 11.0%, and 42.0% of them mentioned that they preferred colonoscopy, double-contrast barium enema (DCBE), and FIT, respectively, if these screening programs were free of charge. On the other hand, 25.5%, 12.5%, and 62.0% of the participants would prefer colonoscopy, DCBE, and FIT, respectively, if they had to pay out of their own pockets. The acceptable copayment amounts from their own pockets were US$93, US$46, and US$3 (3,000, 1,500, and 100 THB) from the full costs of US$186, US$93, and US$3 (6,000, 3,000, and 100 THB) for colonoscopy, DCBE, and FIT, respectively (Supplementary Table 7E). However, 71 participants (17.8%) refused to undergo any CRC screening with various reasons, such as no symptom, busy, and afraid to know the results.

Table 3. Demographic data of participants.

Parameters	Total (n = 400)	Individuals with symptoms (n = 200)	Individuals without symptoms (n = 200)	p-Value
Age (years) (mean ± SD)	62.4 ± 6.4	61.6 ± 6.8	63.2 ± 5.9	.010
Female (n, %)	246 (61.5)	122 (61.0)	124 (62.0)	.837
Comorbidities (n, %)
Hypertension	203 (51.0)	104 (52.0)	99 (50.0)	.689
Dyslipidemia	123 (31.0)	51 (26.0)	72 (36.0)	.030
Diabetes mellitus	97 (24.0)	47 (24.0)	50 (25.0)	.816
Coronary artery disease	24 (6.0)	12 (6.0)	12 (6.0)	1.000
Education (grades) (mean ± SD)	10.5 ± 5.3	10.1 ± 5.2	11.0 ± 5.5	.087
Retired (n, %)	229 (57.3)	110 (55.0)	119 (59.5)	.363
Monthly income (median, IQR)
(US$)	495 (217–929)	464 (190–929)	619 (279–929)	.810
(THB)	16,000 (7,000–30,000)	15,000 (6,000–30,000)	20,000 (9,000–30,000)
VAS self-health rating (0–10) (mean ± SD)	6.4 ± 1.4	6.2 ± 1.5	6.6 ± 1.3	.007
Health insurance schemes (n, %)				<.001
CSMBS	251 (63.0)	106 (53.0)	145 (73.0)
OOP	68 (17.0)	36 (18.0)	32 (16.0)
UCS	57 (14.0)	36 (18.0)	21 (20.0)
SSS	24 (6.0)	22 (11.0)	2 (1.0)
First degree family history of CRC (n, %)	24 (6.0)	12 (6.0)	12 (6.0)	1.000
Known relatives/friends had CRC (n, %)	144 (36.0)	76 (38.0)	68 (34.0)	.405
Fear of CRC (n, %)				.899
Very much	96 (24.0)	76 (38.0)	86 (43.0)
Minimal	142 (35.5)	80 (40.0)	62 (31.0)
No	162 (40.5)	44 (22.0)	52 (26.0)
Aware of existence of CRC (n, %)	214 (53.5)	115 (57.5)	99 (49.5)	.109
Aware of existence of CRC screening (n, %)	94 (23.5)	59 (29.5)	35 (17.5)	.005
Preferred choice when free of charge (n, %)				.010
Colonoscopy	189 (47.0)	82 (41.0)	107 (54.0)
Double-contrast barium enema	43 (11.0)	29 (14.0)	14 (7.0)
Fecal immunochemical test	168 (42.0)	89 (45.0)	79 (39.0)
Preferred choice when OOP is needed (n, %)				.182
Colonoscopy	102 (25.5)	43 (22.0)	59 (30.0)
Double-contrast barium enema	50 (12.5)	27 (13.0)	23 (11.0)
Fecal immunochemical test	248 (62.0)	130 (65.0)	118 (59.0)
Willingness-to-pay OOP
Colonoscopy (median, IQR)
(US$)	93 (46–93)	93 (31–93)	93 (62–93)	.032
(THB)	3,000 (1,500–3,000)	3,000 (1,000–3,000)	3,000 (2,000–3,000)
Double-contrast barium enema (median, IQR)
(US$)	46 (31–46)	46 (15–46)	46 (31–62)	.178
(THB)	1,500 (1,000–1,500)	1,500 (500–1,500)	1,500 (1,000–2,000)
Fecal immunochemical test (median, IQR)
(US$)	3 (3–3)	3 (3–3)	3 (3–3)	.233
(THB)	100 (100–100)	100 (100–100)	100 (100–100)

Abbreviations. CSMBS, Civil Servant Medical Benefit Scheme (government-based employee scheme); IQR, interquartile range; OOP, out-of-pocket; SD, standard deviation; SSS, Social Security Scheme (private employee scheme); UCS, Universal Coverage Scheme (scheme for non-government or private employee); VAS, Visual Analogue Scale.

Table 4 shows the estimated coefficients of all attributes. The positive value of the constant in the model indicated that the participants preferred the CRC screening over no screening. The positive sign of the coefficients implied that the participants preferred the attribute levels and the negative sign implied the opposite. Overall, the participants preferred the screening with high risk reduction of CRC-related mortality, no complication, less frequency interval, less bowel preparation, and lower cost.

Table 4. Multinomial logit model.

Attributes	Total			Symptomatic			Asymptomatic
	β-coefficients (95% CI)	z	p-Value	β-coefficients (95% CI)	z	p-Value	β-coefficients (95% CI)	z	p-Value
CONSTANT	0.31296*** (0.12977, 0.49614)	3.35	.001	0.35368*** (0.08932, 0.61804)	2.62	.009	0.26209** (0.00573, 0.51844)	2.00	.045
Pain	−0.07518 (−0.20748, 0.05711)	−1.11	.265	−0.08993 (−0.28164, 0.10178)	−0.92	.358	−0.04979 (−0.23466, 0.13508)	−0.53	.598
Risk reduction of CRC mortality	0.02069*** (0.01861, 0.02278)	19.44	<.001	0.02171*** (0.01871, 0.02472)	14.16	<.001	0.01963*** (0.01670, 0.02257)	13.11	<.001
Risk of complications	−0.27750*** (−0.42705, −0.12796)	−3.64	<.001	−0.26945** (−0.48589, −0.05300)	−2.44	.015	−0.26604** (−0.47466, −0.05743)	−2.50	.012
Interval 1 year	−0.28600*** (−0.41264, −0.15936)	−4.43	<.001	−0.18926** (−0.37105, −0.00747)	−2.04	.041	−0.38456*** (−0.56285, −0.20628)	−4.23	<.001
Interval 5 years	0.28381*** (0.17725, 0.39036)	5.22	<.001	0.23421*** (0.08064, 0.38777)	2.99	.003	0.33091*** (0.18170, 0.48011)	4.35	<.001
Interval 10 years	0.16388*** (0.05447, 0.27328)	2.94	.003	0.07980 (−0.07821, 0.23781)	0.99	.322	0.23067*** (0.07679, 0.38454)	2.94	.003
Minimal bowel preparation	−0.05585 (−0.14755, 0.03585)	−1.19	.233	−0.09926 (−0.23200, 0.03347)	−1.47	.143	−0.02186 (−0.15035, 0.10664)	−0.33	.739
Intensive bowel preparation	−0.22938*** (−0.32249, −0.13626)	−4.83	<.001	−0.24000*** (−0.37617, −0.10384)	−3.45	.001	−0.22451*** (−0.35436, −0.09466)	−3.39	.001
Out-of-pocket cost	−0.00019*** (−0.00022, −0.00017)	−14.13	<.001	−0.00021*** (−0.00025, −0.00017)	−10.79	<.001	−0.00017*** (−0.00021, −0.00013)	−9.01	<.001

Abbreviations. CI, confidence interval; CRC, colorectal cancer.

***, ** Indicate Significance at 1%, 5% level, respectively; z referred to z-statistic based on standard normal distribution.

The participants were willing to pay US$46 (1,500 THB) more for 5-year interval and US$3 (100 THB) more for every 1% increased risk reduction of CRC-related mortality as compared to no risk reduction of CRC-related mortality, whereas they were willing to pay US$45 (1,454 THB) less for complication as compared to no complication and US$38 (1,227 THB) less for full bowel preparation as compared to no preparation (Supplementary Table 5E). All statistical different level of coefficient of each attribute was consistent among participants with and without GI symptoms, except 10-year interval that was statistically significant only among asymptomatic group (p < .05).

Table 5 shows the participant’s WTP for each screening method. The participants were willing to pay US$51 (1,647 THB) for the CRC screening as compared to no screening. After considering the attributes and levels of all CRC screening methods, they were willing to pay US$189, US$142, US$183, US$154, and US$251 (6,105, 4,587, 5,911, 4,974 and 8,107 THB) for colonoscopy, flexible sigmoidoscopy (FS), DCBE, computed tomographic colonography (CTC), and FIT, respectively (Table 5). The estimated preferred choices of colonoscopy, FS, DCBE, CTC, FIT, and no screening were 11.4%, 11.4%, 14.6%, 9.2%, 38.2%, and 15.3%, respectively (Supplementary Table 2E). Among the participants with and without GI symptoms, FIT was consistently found to be the most preferred screening method for both groups.

Table 5. Analysis of willingness-to-pay.

Willingness-to-pay (US$ and THB) (95% CI)	Individuals with symptoms (n = 200)	Individuals without symptoms (n = 200)	Total (n = 400)
Colonoscopy
US$	172 (−13 to 361)	211 (−15 to 444)	189 (49 to 333)
THB	5,556 (−420 to 11,660)	6,815 (−485 to 14,341)	6,105 (1,583 to 10,756)
Flexible sigmoidoscopy
US$	125 (−40 to 291)	165 (−34 to 370)	142 (17 to 269)
THB	4,038 (−1,292 to 9,399)	5,330 (−1,098 to 11,951)	4,587 (549 to 8,689)
Double-contrast barium enema
US$	163 (−9 to 339)	209 (0 to 425)	183 (52 to 317)
THB	5,265 (−291 to 10,950)	6,751 (0 to 13,728)	5,911 (1,680 to 10,239)
Computed tomographic colonography
US$	143 (−31 to 319)	171 (−41 to 389)	154 (23 to 289)
THB	4,619 (−1,001 to 10,304)	5,523 (−1,324 to 12,565)	4,974 (743 to 9,335)
Fecal immunochemical test
US$	265 (202 to 336)	235 (154 to 326)	251 (201 to 304)
THB	8,560 (6,525 to 10,853)	7,591 (4,974 to 10,530)	8,107 (6,492 to 9,819)

Abbreviation. CI, confidence interval.

Discussion

Our findings showed the consistency between the results and our hypothesis based on clinical knowledge. Moreover, the beta coefficient of each attribute can represent the different magnitude of influential level affecting to the screening decision. The participants preferred screenings with high risk reduction of CRC-related mortality, no complication, 5-year interval, and lower cost. The pain attribute showed insignificant effect in our results. One of the reasons for this effect was that pain could be a subjective attribute and participants might not be able to distinguish between mild and no pain levels. FIT yielded the highest WTP value of around 1.3–1.8 times higher WTP when compared to other screening tests, indicating that it was the most preferred screening method. This finding is also evident by the high estimated preferred choices of almost 40% with the highest WTP value.

Our findings were similar to literature in the field of acceptability and individual preferences for the colorectal screening²⁵. In the Netherlands, two DCEs were conducted among screening-naive participants and previously screened participants, aged 50–75 years. The results showed that pain, risk of complications, screening location, preparation, duration of procedure, screening interval, and risk reduction of CRC-related death proved to significantly influence the participants’ preferences²⁵. In contrast, another study in the Netherlands showed that the study participants equally preferred 5-yearly flexible sigmoidoscopy and 10-yearly colonoscopy. They favored endoscopic strategies to annual FIT due to the more favorable risk reduction of CRC-related mortality²⁵. In Australia, the study participants preferred colonoscopy over CTC²¹. In the US, the study participants preferred shorter travel, rewards or small copayments, stool testing, and greater coverage of follow-up costs.²³ However, previous DCE studies were conducted in developed countries, which have different economies and healthcare systems from LMICS.

The estimated preferred choices from the DCE results were consistent with those results from the direct questions of the study. The participants preferred FIT the most of all three screening methods if they needed to pay out-of-pocket. Moreover, the significant cost attribute from the DCE results confirmed the results from the direct questions of the study indicating when the participants would prefer the colonoscopy, which is the most expensive method, if they did not need to pay out-of-pocket. The proportions of those individuals who decided to undergo no screening from both DCE and direct questions were also similar (Supplementary Table 6E).

Our study findings were similar to a previous survey study in Thailand that was undertaken to assess the acceptance of FIT and colonoscopy in participants aged 50–69 during their visits to a primary care unit²⁷. The FIT was accepted by 74.1%, as compared to 55.6% for colonoscopy. The FIT was preferred for its simplicity and non-invasiveness. No symptoms, unwilling to screen, healthy, too busy, and anxious about diagnosis were reasons for refusing to screen²⁷. The measurement of acceptance in that study using the questionnaire was valuable but it did not provide details, including their preferences on the attributes of the screening methods. On the other hand, our study is the first DCE study that rigorously provided stated preference in CRC screening in Thailand. Our study findings can provide clinicians and policy makers some insights and opportunities to promote certain CRC screening methods by targeting favorable or concerning attributes of these methods. This would enhance the likelihood of success of the CRC screening policy in Thailand.

Our study also revealed that the absence of awareness of CRC and CRC screening was not uncommon. This absence could be a major barrier of CRC screening participation. According to the study results, only about 20% of participants knew that CRC screening programs were available in the Thai healthcare system. To improve the screening rate and compliance to the screening programs, we should improve understanding and awareness on the disease and intervention among the population at risk as well as remind them to repeat the test as recommended¹¹. In addition, the cost of screening is very important for the participants. Thai government should financially support CRC screening since the people at risk in the country might not take the screening test if they needed to pay out of their own pockets. The government should also portray all subsequent costs, such as cost of treatments, cost of complications, cost of follow-up, and other indirect costs, as a result of the screening in order to lessen the level of uncertainty for the people at risk. In addition, in terms of both human and monetary resource limitation, prolonging the interval between each screening would result in much less burden, especially in LMICs. Nevertheless, the clinical benefit of the screening strategy, itself, may also decrease.

While the incorporation of the preferences of patients or relevant population who are directly affected by technology in the development of value assessment policy has been the center of healthcare decision making in the developed countries, such as US¹¹ and European countries⁴¹, it was rarely found in Thailand and other LMICs. The findings of this study would complement the recently published findings from the cost-effectiveness and budget impact analyses of the CRC screening programs which indicated that the colonoscopy offered best value for money of CRC screenings and that the FIT is a more affordable screening method in Thailand²⁸. Our study findings also indicated that the FIT was the most preferred screening method. In other words, the study findings would allow healthcare policy makers to consider or incorporate the preferences and willingness-to-pay of the population at risk in the HTA process or in policy development for the CRC screenings in Thailand. For instance, if the policy makers decide to support the colonoscopy due to its best value, then they need to better educate the population at risk about those important attributes of the colonoscopy, including financial assistance, instead of emphasizing the value for money alone. Other LMICs could also adopt the study findings or use a similar approach that includes both QALY-based and preference-based value assessments to develop their policies. Because not only the availability of screening tests in the countries, the acceptability of the population is also very important to show its effectiveness and policy success, especially in limited-resource countries.

There are various limitations in this study. First, although the study was conducted in Thailand, the study findings should be interpreted with caution in terms of generalizability for the entire Thai population. This study took place at the waiting areas of a hospital and the participants might have different characteristics. For instance, the study participants could be the persons who likely had more chronic diseases and more concern about their health. They could possibly access the hospital services more than other people did. These differences could affect their preferences and WTP. Similarly, although the findings could possibly be applied to other LMICs, they should be cautious since those countries might have different sociodemographic charteristics leading to different preferences and WTP. Second, this study is a stated preference, which may be different from revealed preference⁴². The revealed preferences should be examined after implementing the CRC screening programs. Third, the MNL, which was used to develop the utility model in this study, is under restricted assumptions, which did not allow this study to examine preference heterogeneity of the population at risk. However, the study findings could provide prior parameters for other advance models of future studies. In addition, one of the possible heterogeneities, which was the preferences of individuals with and without GI symptoms, was examined in this study. Another limitation was that some estimated WTPs were higher than the range of the cost attributes. Therefore, these results should be cautiously considered and investigated further.

Conclusions

The risk reduction of CRC-related mortality, complication, screening interval, and bowel preparation were important to Thai adults when they considered the CRC screening. They preferred the FIT over the other screening tests. Health policy makers could incorporate the preferences of population at risk to formulate the policy that improves the success rate of CRC screening campaign and can subsequently be implemented as a national health policy. They can also complement QALY-based value assessment as such cost-effectiveness analysis in HTA. Therefore, Thai population at risk would receive preferred CRC screening, early diagnosis of the precancerous lesions, and early treatment.

Transparency

Declaration of funding

The authors received no funding for this work.

Declaration of financial/other interests

Pochamana Phisalprapa (PP), Surachat Ngorsuraches (SN), Tanatape Wanishayakorn (TW), Chayanis Kositamongkol (CK), Siripen Supakankunti (SS), and Nathorn Chaiyakunapruk (NC) declare no potential competing interest with respect to the research, organization, and publication of this work.

Author contributions

PP and CK participated in the study concept and design, data acquisition, data analysis, data interpretation, manuscript drafting, critical revision of the manuscript, and the final approval of the manuscript. SN, TW, SS, and NC participated in the study concept and design, data interpretation, manuscript drafting, critical revision of the manuscript, and the final approval of the manuscript.

Previous presentations

This work was presented at International Society for Pharmacoeconomics and Outcomes Research (ISPOR) 23rd Annual International Meeting in Baltimore, USA on 23 May 2018.

Ethics approval and consent to participate

The protocol for this study was approved by the Siriraj Institutional Review Board (SiRB) (COA no. Si 363/2017). Written consent was obtained from all participants in the study.

Acknowledgements

The authors would like to thank Ph.D. program in Economics, Faculty of Economics, Chulalongkorn University for academic supports, Malee Anakul, research assistants, for collecting the data, and Madeline Kelsey Brendle for editorial support in the manuscript.

Notes

i. Ngene is a registered trademark of ChoiceMetrics Pty Ltd., NSW, Australia.

ii. Nlogit is a registered trademark of Econometric Software, Inc., NY, USA.