Organizational Predictors of Colonoscopy Follow-up for Positive Fecal Occult Blood Test Results: An Observational Study

Colorectal cancer is the third most common cancer and the third leading cause of cancer-related death among men and women in the United States (1). The best known defense against colorectal cancer is early detection and prevention through routine screening. Current guidelines endorse multiple colorectal cancer screening methods (2–4), but fecal occult blood testing (FOBT) and colonoscopy are the most widely used (5). Two of the largest integrated health care systems in the United States [Kaiser Permanente and the Veterans Health Administration (VHA)] have achieved high colorectal cancer screening rates using screening programs emphasizing FOBT (6, 7). Although randomized controlled trials have demonstrated that FOBT can be a highly efficacious screening method if FOBT-positive (FOBT⁺) results are followed by diagnostic colonoscopy (8–10), many FOBT-based screening programs document challenges assuring that FOBT⁺ results receive follow-up colonoscopy in a timely manner (11–14). Proportions of FOBT⁺ cases failing to receive follow-up colonoscopy reported in prior studies range from 35% to 63% (11–14), and the median waiting times from FOBT⁺ to colonoscopy range from 105 to 202 days (11–13, 15).

Both the VHA and the Canadian Association of Gastroenterology Wait Time Consensus Group recommend performing a colonoscopy within 60 days of FOBT⁺ results (16, 17). However, recent data from the VHA documenting that 50% of FOBT⁺ cases fail to receive follow-up colonoscopy within this window (18), and from a Canadian survey of gastroenterologists documenting an average waiting time from FOBT⁺ results to colonoscopy of 105 days (15), suggest significant gaps remain in assuring timely follow-up. Closing these gaps will require identifying modifiable contributors to persistent FOBT⁺ follow-up delays.

Most prior studies examining contributors to FOBT⁺ follow-up have focused on nonmodifiable individual-level factors (19, 20), identifying significant associations with patient age (21–24), gender (13, 23, 25), race (24), comorbidity (22), personal history of bowel disease (26), family history of colorectal cancer (27), and recent colonoscopy (24). A few have identified modifiable individual-level contributors, including patient fears and worries about colorectal cancer (26, 27), and provider awareness of guidelines (28, 29) and intentions to order diagnostic testing for FOBT⁺ results (25). However, very few prior studies have examined the contribution of modifiable organizational-level factors to FOBT⁺ follow-up patterns (20), despite growing recognition that understanding the role that the care environment plays in assuring quality is essential to designing effective interventions and making further improvements in cancer care (30, 31).

We conducted a study to assess the contribution of modifiable organizational-level factors to FOBT⁺ follow-up rates. We hypothesized that higher follow-up rates would be associated with: (i) organizational structures designed to facilitate quality improvement (i.e., leadership support, resource alignment, feedback, and incentives), and (ii) organizational processes that control system-wide demand for colonoscopy (i.e., “demand efficiency” processes), minimize wasted appointments and the number of steps required to complete a colonoscopy (i.e., “supply efficiency” processes), and address patient barriers to colonoscopy completion (i.e., “patient-centered” processes). We assessed the contribution of these organizational-level factors while controlling for individual-level factors demonstrated to be associated with FOBT⁺ follow-up in prior studies.

We identified a cohort of patients who had outpatient FOBT⁺ results from a VHA facility between August 2009 and March 2011 (1 year prior and 6 months after the start date for the organizational survey, described below) and followed them until September 2011 (6 months after the last FOBT date) for completion of follow-up colonoscopy. To identify patients with FOBT⁺ results, we identified from VHA laboratory records all outpatient FOBT procedures performed at VHA facilities during the sample accrual period, using the codes provided in Supplementary Material S1. We then defined FOBT⁺ cases as any individual card test with a positive result, or any multiple card series with one or more cards with positive results. If an eligible patient had more than one FOBT⁺ result in the sample selection window, the first FOBT⁺ result was selected for the sample. We then excluded patients if they did not receive their FOBT⁺ results from one of 125 VHA facilities that conducted at least 1,400 FOBTs in 2009; were of age <18 years or >100 years at the time the FOBT⁺ result was recorded; had a prior diagnosis of colorectal cancer in VHA medical records; or received their FOBT⁺ from a VHA community-based outpatient clinic that refers less than 70% of colonoscopies to one of the 125 VHA facilities included in the sampling frame, leaving 86,926 eligible FOBT⁺ patients available for analysis. We linked this patient sample to facility-level data on organizational structures and processes obtained from Web-based surveys (described previously in ref. 32 and in Supplementary Material S2) administered to the chiefs of primary care (73% response rate) and gastroenterology (81% response rate) beginning in August 2010, yielding 74,014 patients from 98 facilities responding to one or both surveys (Fig. 1). We excluded 43 facilities completing <1,400 FOBT procedures in 2009 to ensure an adequate sample (≥100) of FOBT⁺ patients from each facility was available for estimating the association between organizational factors and follow-up rates.

Our conceptualization of key organizational-level contributors to FOBT⁺ follow-up is informed by Donabedian's framework for understanding the quality of care (33), and prior research documenting the association among specific organizational structures, organizational processes, individual-level factors, and quality outcomes. Organizational structures associated with quality outcomes in prior studies include leadership support (34, 35), personal and frequent feedback (36, 37), incentives (38), and resource alignment to improvement goals (refs. 34, 35, 39; such as tracking systems, ref. 40; and quality improvement training, ref. 14). Organizational processes refer to approaches used to complete each step required to assure FOBT⁺ follow-up (i.e., notification, referral, scheduling, and patient education). For our analysis, we categorized organizational processes into three groups: (i) those that control system-wide demand for colonoscopy, which we refer to as “demand efficiency” processes; (ii) those that minimize wasted appointments and the number of steps required to complete a colonoscopy, which we refer to as “supply efficiency” processes; and (iii) those that address patient barriers to colonoscopy completion, which we refer to as “patient-centered” processes. Individual-level factors we control for in our hypothesis tests related to organizational-level factors include characteristics of patients and FOBT procedures found to be associated with FOBT⁺ follow-up rates in prior studies (i.e., age, race, residence, comorbidities, personal history of polyps, and ordering provider characteristics).

Our primary outcome was follow-up colonoscopy completion, identified from VHA administrative records using the codes in Supplementary Material S3. We separately examined correlates of colonoscopy completion within 60 days (the VHA recommended follow-up interval) and 6 months.

Table 1 provides the survey question wording, response options, and coding for analysis for all organizational predictors we examined.

We measured leadership support using two items: (i) a question from the primary care survey asking the extent to which “not a priority to leadership” is a barrier to providing timely FOBT⁺ follow-up, and (ii) an identical question from the gastroenterology survey. We measured resource alignment with two items: (i) “tracking,” a question on the primary care chief survey about how frequently their program tracks what happens to patients with FOBT⁺ results, and (ii) a question on the gastroenterology chief survey asking the extent to which “colonoscopy appointment availability” is a barrier to providing timely FOBT⁺ follow-up. Feedback was assessed with two measures: (i) “primary care feedback,” which combined two measures on the type and frequency of feedback (see Table 1) into a single measure, and (ii) “gastroenterology feedback,” which combined similar measures from the gastroenterology survey. Incentives were assessed with two items: (i) “primary care incentives,” and (ii) “gastroenterology incentives,” both of which asked chiefs “which of the following do (providers in your primary care program/staff in your gastroenterology program) receive for their performance on assuring timely follow-up of positive FOBT results” (see Table 1 for response options and coding).

Measures of demand efficiency processes included: (i) number of contraindications the facility's colorectal cancer screening clinical reminder asks about; (ii) information on the facility's colonoscopy consult template includes contraindications; and (iii) the typical surveillance interval for patients with 1–2 adenomas <1 cm at the facility is guideline-consistent (i.e., 5–10 years; ref. 41). Measures of supply efficiency processes included: (i) Gastroenterology providers are directly notified of FOBT⁺ cases (either by laboratory or by gastroenterology staff who take responsibility for identifying FOBT⁺ cases); (ii) patient colonoscopy prep instruction does not require a separate appointment; (iii) a pre-op appointment is not required for colonoscopy; and (iv) overbooking is used to minimize wasted appointments. Measures of patient-centered processes included: (i) patient notification of FOBT⁺ results includes phone contact; (ii) colonoscopy appointment times are negotiated (using a scheduler or letter requesting the patient call the clinic to set up an appointment) rather than preassigned to patients in a mailed letter; (iii) patient prep instruction procedures include opportunities for questions (i.e., some verbal instruction); and (iv) patients receive colonoscopy appointment reminders that include prep instructions.

We used VHA administrative data to identify the following individual-level factors controlled for in our predictive models: age (<50, 50–64, 65–84, or ≥85); race (non–Hispanic white, Hispanic, African American, American Indian, Asian or Pacific Islander, or Unknown); residence (urban or rural); drive time to the nearest VHA specialty care facility (≤60 or >60 minutes); Charlson comorbidity score for the 1 year before the FOBT⁺ result; mental health diagnoses (psychiatric only, substance abuse only, dual diagnosis, or none); personal history of colorectal polyps or benign neoplasms (ICD-9 211.3-4, 569.0, v12.72); whether the FOBT was ordered by their primary care provider; what type of provider ordered their FOBT (physician, nurse practitioner or physician assistant, resident, nurse, or other staff); and what type of facility ordered their FOBT (VHA specialty care facility or community-based outpatient clinic).

Because the limited number of facilities in the sample precluded the simultaneous inclusion of a large number of facility-level predictors, we pursued the following steps to select predictors for inclusion in the final model. We initially fit separate bivariate hierarchical logistic regression models (with random effects for facility of care) for each organizational structure and process measure. We then included in a base multivariable model all measures with P < 0.10, or with more than a 5% difference between model estimated completion rates (among the levels of a categorical measure or between the mean and one standard deviation shift from the mean for a continuous measure). We also included in the base model month of FOBT⁺ result and any patient-level predictors associated with colonoscopy completion within the respective timeframe (see Supplementary Table S1 for bivariate estimates derived from this step). We then reduced the number of predictors in this model in a stepwise fashion, retaining explanatory measures with P < 0.10. Using this final multivariable model, we constructed model-based odds ratios (OR) and least square mean completion rate estimates (using the observed marginal distributions of the other covariates) for each of the organizational structure and process measures included in the model. To explore the impact of potentially valid reasons for not completing a follow-up colonoscopy at a VHA facility on our estimates, we fit a final set of models excluding patients who may not have been appropriate for colorectal cancer screening (i.e., age <45 or >85, with documentation of limited life expectancy in the medical record, or a colonoscopy in the prior 10 years), and treating patients who refused colonoscopy or chose to pursue colonoscopy in the private sector as having adequate follow-up. Supplementary Material S4 provides details on how we estimated the prevalence of each of these reasons for not completing a colonoscopy.

The study was reviewed and approved by the Institutional Review Boards at the Minneapolis VA Medical Center (Minneapolis, MN; approval September 15, 2009), and the Boston VA Medical Center (Boston, MA; approval February 10, 2010).

Patients included in the analysis were primarily non–Hispanic white (65%) married (53%) men (96%) older than 50 years (95%; Table 2). Roughly half (52%) lived in urban areas, and most (61%) lived less than 60 minutes from a VHA medical center providing specialty services. A total of 20% had a personal history of colorectal polyps or benign neoplasms, 68% had a mental health or substance abuse diagnosis, and the mean Charlson comorbidity score was 1.9. Most participants (58%) had their FOBT procedures ordered at a VHA hospital, by their primary care provider (70%), and most ordering providers were physicians (72%).

Few facilities cited lack of leadership support as a barrier to improving FOBT⁺ follow-up rates, but 38 (49%) considered colonoscopy appointment availability a barrier (Table 3). A total of 36% reported tracking what happens to patients with FOBT⁺ results on a weekly or more frequent basis, but only 11% reported providing individual, frequent feedback to primary care staff about their FOBT⁺ referral practices. A higher percentage (22%) reported providing verbal, frequent feedback to gastroenterology staff about colonoscopy follow-up rates. The majority reported no incentives for primary care (56%) or gastroenterology staff (64%) tied to FOBT⁺ follow-up performance. About half (48%) reported including some information on contraindications on their colorectal cancer screening clinical reminder, and 47% reported including information on contraindications on their colonoscopy consult template. The modal surveillance interval for patients with 1–2 adenomas <1 cm was 5 years (84%). Most (70%) relied on primary care to notify gastroenterology of FOBT⁺ cases, and most (70%) did not require a separate appointment for colonoscopy prep instruction. The majority (62%) did not require a pre-op appointment for colonoscopy, and 54% reported using overbooking to minimize wasted colonoscopy appointments. Most (67%) used patient notification procedures that included some phone contact, but scheduling procedures were highly variable. The approach used to instruct patients about colonoscopy preparation was highly variable, with 44% using written methods only, 23% using verbal methods delivered by phone or an individual appointment, and 32% using group appointments or some other method involving both written and verbal instruction. Finally, most (69%) did not review prep procedures in their colonoscopy appointment reminders.

The cumulative proportion of patients with FOBT⁺ results receiving a colonoscopy at a VHA facility within 60 days was 30% (range, 10%–57% across facilities). Organizational structures significantly associated with 60-day follow-up rates in the multivariable model included: colonoscopy appointment availability is not a key barrier [OR, 1.43; confidence interval (CI), 1.09–1.90; P = 0.01]; and monthly or more frequent primary care feedback (OR, 1.79; CI, 1.02–3.16; P = 0.04; Table 4). Organizational processes significantly associated with 60-day follow-up rates in the multivariable model included: colonoscopy consult information does not ask about indication or contraindications for colonoscopy (OR, 1.48; CI, 1.07–2.05; P = 0.02), or asks about indication only (OR, 1.49; CI, 1.10–2.02; P = 0.01; a finding counter to our demand efficiency hypothesis), and gastroenterology notification directly of FOBT⁺ results by gastroenterology staff (OR, 1.85; CI, 1.17–2.91; P = 0.01). No patient-centered process measures were significantly associated with the 60-day outcomes in the adjusted model. After excluding potentially inappropriate FOBTs, and treating refusals and private sector colonoscopies completed after the FOBT⁺ as adequately followed up (last two columns of Table 4), the estimated follow-up rates increase markedly, and the OR estimates for colonoscopy appointment availability and gastroenterology notification remain significant. However, the OR estimates for primary care feedback, and colonoscopy consult information attenuate and are no longer statistically significant.

The cumulative proportion of patients with FOBT⁺ results receiving a colonoscopy within 6 months was 49% (range, 30%–70% across facilities). No organizational structures were significantly associated with the 6-month outcome in the multivariable model. Organizational processes significantly associated with 6-month follow-up rates in the multivariable model included using an adenoma surveillance interval of at least 5 years (5 years OR, 1.32; CI, 1.02–1.71; P = 0.04; 7–10 years OR, 1.57; CI, 1.11–2.20; P = 0.01); having gastroenterology notification directly from laboratory about FOBT⁺ cases (OR, 1.25; CI, 1.06–1.47; P = 0.008); and using group appointments or other combined verbal and written methods to provide patient colonoscopy prep instruction (OR, 1.48; CI, 1.22–1.79; P = 0.05; Table 5). After excluding potentially inappropriate FOBTs, and treating refusals and private sector colonoscopies completed after the FOBT⁺ as adequately followed up (last two columns of Table 5), the estimated follow-up rates increase to more than 67% in all subgroups, and only the estimate of group appointment for colonoscopy prep instructions remained significant (OR, 1.50; CI, 1.16–1.95; P = 0.003).

Consistent with our hypothesis that organizational structures supporting quality improvement would be positively associated with follow-up rates, we found that adequacy of colonoscopy appointment availability, and providing primary care providers with individual, frequent feedback about the timeliness of FOBT⁺ referrals were positively associated with receiving follow-up colonoscopy within 60 days of an FOBT⁺. These findings are consistent with prior research documenting that limited colonoscopy capacity is the most common barrier to reducing FOBT⁺ follow-up delay (40), and that individual, frequent feedback can improve adherence to clinical practice guidelines (42, 43). However, the fact that no organizational structures were significantly associated with 6-month follow-up rates suggests that these factors have more influence on how quickly, rather than whether, patients with FOBT⁺ results receive colonoscopy follow-up.

We found partial support for our hypotheses that higher follow-up rates would be associated with organizational processes enhancing demand efficiency, supply efficiency, and patient-centered processes. Specifically, our findings suggest that: using surveillance intervals for low-risk adenomas that are not more aggressive than recommended by guidelines (ref. 41; a demand efficiency measure); assigning responsibility for identifying FOBT⁺ cases to laboratory or gastroenterology staff (a supply efficiency measure); and using group and other combined verbal and written colonoscopy prep instruction processes (a patient-centered process) are positively associated with overall follow-up rates.

One organizational process measure (gastroenterology providers are notified by laboratory or gastroenterology staff about FOBT⁺ results) was significantly associated with both 60-day and 6-month follow-up rates. These results are consistent with findings from a prior randomized trial conducted in four VHA facilities, which found 30-day, 90-day, and 6-month follow-up rates improved significantly (by 9%–31%; P < 0.03) in facilities that implemented an electronic intervention to directly notify gastroenterology staff of FOBT⁺ results, but did not significantly change in the usual care comparison facilities (44).

Counter to our hypothesis that organizational processes designed to reduce unnecessary demand for colonoscopy would be positively associated with follow-up rates, we found facilities that asked about contraindications on colonoscopy consult templates had lower rather than higher 60-day follow-up rates. This finding might be explained by the fact that patients with documented limited life expectancy, recent colonoscopy, and refusal to complete follow-up colonoscopy were not initially excluded from our sample. Indeed, after excluding these cases from the analysis, the association of consult template characteristics with follow-up rates was no longer statistically significant. Thus, including information on contraindications on the colonoscopy consult template may identify individuals who should not have been screened, and some of these individuals appropriately do not have follow-up colonoscopy.

We hypothesized that colonoscopy prep instruction processes that involve some verbal instruction would be associated with higher follow-up rates than methods that involved only written instruction because verbal instruction processes would provide more opportunities to address patient questions. However, our findings suggest that some forms of verbal instruction (i.e., phone and individual appointments) were associated with lower 6-month follow-up rates than written only instruction methods. Because we did not collect information on the specific content of the prep instruction provided, we can only speculate about why group preparation instruction and other combined verbal and written methods were superior to verbal phone and individual appointment instruction. One possibility is that group prep appointments and other combined methods may use a more structured approach than other verbal instruction methods, and so are more likely to encourage patients to identify and clarify aspects of the preparation they do not understand. Alternatively, the group/peer setting and other combined approaches may prompt greater patient engagement. The resulting enhanced clarification and/or engagement may increase the proportion of patients that attend their scheduled colonoscopy appointment and present with adequate bowel preparation, thereby reducing delays associated with needing to reschedule colonoscopy appointments. A final possibility is that instruction approaches that require patients to complete a group class or other formal instruction before scheduling a colonoscopy lead to self-selection of individuals that are more likely to adhere to their colonoscopy appointment. All of these explanations are consistent with findings from one prior study, which found that patients participating in a nurse-led group colonoscopy prep education program had higher colonoscopy completion rates and lower cancellation rates due to poor bowel preparation than patients who received an educational brochure only (45). Our finding that other measures of patient-centered processes (phone results notification, negotiated appointment scheduling, and appointment reminders that review prep instructions) were not associated with follow-up rates contrasts with previous studies attributing high endoscopy attendance to patient-centered processes such as education (46) and reminder systems (47, 48), may be unique to this patient population trained in the hierarchical traditions of the military, and may not generalize to other health care settings.

This study has a number of strengths, including the large sample size of patients and medical facilities, the rigorous methodology used to adjust our estimates for reasons a colonoscopy was not completed, and the fact that it identifies several modifiable organizational predictors of FOBT⁺ follow-up rates. However, our findings should be qualified by several limitations. First, we may be underestimating actual follow-up rates because some patients with FOBT⁺ results may have pursued colonoscopy outside of the VHA. Indeed secondary analyses we conducted on this cohort suggest that up to 15% of patients with FOBT⁺ results who did not receive a colonoscopy in VHA within 6 months had documentation in their chart notes that they were pursuing colonoscopy in the private sector. However, sensitivity analyses treating patients with documentation of pursuing colonoscopy in the private sector as adequately followed up did not significantly alter the pattern of associations between organizational factors and follow-up rates reported here. A second possible limitation is that our measures of organizational structures and processes may include some measurement error. Structure and process reports from chiefs were measured at one point in time, in most cases with single-item measures, and may therefore be inaccurate (given that facilities may make periodic adjustments to structures and processes) or insufficiently sensitive (from oversimplification of the underlying processes). Furthermore, lack of variability in our sample forced us to collapse potentially distinct categories for several measures. Future studies should examine whether more detailed measures in more variable facility samples yield different results. In addition, our analysis excluded 12,822 FOBT⁺ patients from 25 facilities with incomplete facility survey data, and 10,806 FOBT⁺ from 43 facilities conducting fewer than 1,400 FOBTs in 2009, which may raise concerns about whether our findings can be generalized to FOBT⁺ patients from other VHA facilities in the sampling frame. However, our previous analysis of the survey data found no significant variation in facility FOBT⁺ follow-up rates or characteristics by survey response status (32), and facilities excluded on the basis of FOBT volume represented not only smaller facilities with FOBT-based screening programs, but also larger facilities with colonoscopy-based screening programs. Finally, the VHA is a unique context, characterized by a predominantly male, low-income population with higher than average comorbidity burden, including high rates of mental health and substance abuse diagnoses. Therefore, our findings may not generalize to other health care contexts. Given that VHA is the largest integrated health care system in the United States, however, our findings have important implications for a substantial population of health providers and consumers in this country.

Despite these limitations, the insights gleaned from this study regarding the role organizational structures and processes can play in assuring patients with FOBT⁺ results receive timely colonoscopy will be helpful in guiding future efforts to improve FOBT⁺ follow-up rates. Specifically, our most robust findings suggest that gastroenterology clinics may be able to significantly increase the proportion of FOBT⁺ results that receive follow-up colonoscopy by assuming responsibility for identifying FOBT⁺ results, and using prep education processes that include both written and verbal information, but to increase the proportion of FOBT⁺ patients that receive follow-up colonoscopy within 60 days, it may be necessary to increase colonoscopy appointment availability. Given that the significant organizational-level predictors of follow-up rates we identified all had modest effects (i.e., resulting in at most 5%–14% differences in follow-up rates), multifaceted strategies designed not only to increase colonoscopy follow-up for FOBT⁺ results, but also to reduce FOBT use in patients who would not complete colonoscopy follow-up due to contraindications or personal preference, may be needed to close remaining gaps. A fruitful area for future research would be the evaluation of such multifaceted strategies.

No potential conflicts of interest were disclosed.

The views expressed in this article are those of the authors and do not necessarily represent the position or policy of the Department of Veterans Affairs or the NIH.

Conception and design: M.R. Partin, D.J. Burgess, J.F. Burgess Jr, D. Haggstrom, A.A. Powell, L.C. Walter, D.B. Nelson

Development of methodology: J.F. Burgess Jr, A. Gravely, S. Nugent, A. Shaukat, D.B. Nelson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.R. Partin, S.E. Lillie, S. Nugent, A.A. Powell

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.R. Partin, D.J. Burgess, A. Gravely, D. Haggstrom, A.A. Powell, A. Shaukat, L.C. Walter, D.B. Nelson

Writing, review, and/or revision of the manuscript: M.R. Partin, D.J. Burgess, J.F. Burgess Jr, A. Gravely, D. Haggstrom, S.E. Lillie, S. Nugent, A.A. Powell, A. Shaukat, L.C. Walter, D.B. Nelson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Gravely, S. Nugent

Study supervision: M.R. Partin

This research was funded by the Department of Veterans Affairs Health Services Research and Development (VA HSR&D) grant IIR 08-334-2 (to M.R. Partin), VA HSR&D Research Career Scientist award RCS 10-185 (to M.R. Partin), VA HSR&D Career Development Award (CDA) #CDA 07-016 (to D. Haggstrom), VA HSR&D CDA #CDA 08-024 (to A.A. Powell), VA Clinical Science Research and Development Career Development Award CDA-2 (to A. Shaukat), and National Institute on Aging at the NIH grant K24AG041180 (to L.C. Walter).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.