Creating a Safety Net Process to Improve Colon Cancer Diagnosis in Patients With Rectal Bleeding

Abstract

Background

Failure to follow up on patients with rectal bleeding is common and may result in a delay in diagnosis of colorectal cancer or in missing high-risk adenomas. The authors’ purpose was to create an electronic patient safety net for those diagnosed with rectal bleeding but who did not have colonoscopy to ensure proper detection of colonic abnormalities, including colon cancer.

Methods

In an integrated health delivery system serving < 4.6 million patients in Southern California, from 2014 to 2019, the authors electronically identified patients with rectal bleeding aged 45 to 80 years but with no recently documented colonoscopy. These cases were reviewed by a gastroenterologist to determine if colonoscopy was appropriate. The physician looked for known documentation as to the cause of rectal bleeding and verified no contraindications to the procedure; if indicated, testing was offered.

Results

Using the authors’ safety net program, 1430 patients with rectal bleeding who needed and completed a colonoscopy were identified. Of those patients, 7.5% had an advanced adenoma or cancer, with a total of 20 cancers, and 34% had findings that warranted more frequent colonoscopy.

Conclusions

The authors designed a safety net system that was able to capture information on patients with rectal bleeding who had not had a colonoscopy and detected in 34% colonic pathology that would have otherwise gone undetected. The program did not require many resources to implement and had the ability to potentially prevent harm from reaching patients whose rectal bleeding did not get prompt workup. Other health systems and practices should consider implementing a similar system.

Introduction

In 2015 the National Academy of Medicine released an important report that highlighted the high frequency of diagnostic error in medical practice. ¹ That report noted that most Americans will experience a diagnostic error in their lifetime, often with serious consequences. It is estimated that about 10% of hospital deaths are associated with a diagnostic error, ¹ and each year in the United States, 5% of adults experienced diagnostic errors in outpatient settings. ²

Reasons for diagnostic errors are complex, ^3–5 and successful methods for reducing them are not entirely known. Some authors ^6–10 have focused on closing the loop on abnormal test results, as this may be something more concrete than trying to mitigate cognitive bias or other causes of diagnostic error. The authors have an interest in this area and created an electronic surveillance system (SureNet) ^11–14 that leverages electronic medical records (EMRs) to identify abnormal test results that do not appear to have adequate follow-up, creating opportunity for an intervention prior to harm reaching the patient. Typically, this involves finding an abnormal test result, defining a desired follow-up, and then looking electronically as to whether the follow-up occurred. If follow-up has not occurred, then the SureNet program intervenes to initiate the needed follow-up prior to harm reaching the patient.

Over time, however, the authors realized that this method could also be applied to symptoms or signs as long as they leave an electronic signal in the EMR. The authors previously reported on closing the loop on follow-up of gross hematuria ¹⁵ and now report on closing the loop on follow-up of rectal bleeding.

Rectal bleeding has many causes; however, it is commonly a symptom of colorectal cancer. ¹⁶ Prompt diagnosis is important because the prognosis of colon cancer is substantially better when diagnosed at an earlier stage. ¹⁷ Failure to promptly work up rectal bleeding is not uncommon, ¹⁸ and missed diagnoses of colon cancer represent a substantial proportion of medical malpractice claims in the United States. ^19,20

In the authors’ previous study of 61 patients who presented with stage III to IV colorectal cancer in the authors’ institution, 11 (18%) presented with rectal bleeding that was not promptly followed up. ²¹ Based on the authors’ findings as well as the literature, the authors decided to create a SureNet program that would identify patients with rectal bleeding who did not appear to have a follow-up colonoscopy to see if neoplastic colonic pathology that has been delayed could be identified.

Methods

This program took place in the Southern California region of Kaiser Permanente, an integrated health care delivery system serving < 4.6 million patients who obtain the vast majority of medical care (including laboratory services, radiology, specialty care such as gastroenterology, pharmacy, inpatient services, and ambulatory care) from within this system. A single multispecialty medical group (Southern California Permanente Medical Group) provides the medical care for all members in this Kaiser Permanente region. All health care encounters within this system are documented within a common EMR system.

Kaiser Permanente’s SureNet program is a regional outpatient safety net that has been active since 2009. ¹¹ SureNet involves a small centralized team using the EMR to systematically identify members who have had inadvertent lapses in care and to intervene before the patient experiences any harm. For this investigation, the SureNet team leveraged the EMRs to screen for members between the ages of 45 and 80 years with a recent diagnosis of rectal bleeding (International Classification of Diseases, Ninth Revision, codes 569.3, and 556.5, International Classification of Diseases, Tenth Revision, codes K62.5 and K51.511) but no colonoscopy referral sent at the time of rectal bleeding diagnosis and no history of colonoscopy in the last 5 years. Included were persons who were diagnosed with rectal bleeding within the previous calendar month, as indicated by an encounter diagnosis or as documented on the Problem List, a list of a patient’s ongoing medical problems, in the EMR. It was elected that Kaiser Permanente’s SureNet team would take responsibility for the data generation, outreach, and tracking component of this program. After identifying qualifying patients, gastroenterologists reviewed the medical records for these patients, and if a colonoscopy was indicated, a nurse engaged the patient and placed a referral for colonoscopy to be cosigned by the primary care physician. Review of existing databases and charts of all patients included in the study was conducted under a protocol approved by the Institutional Review Board of Kaiser Permanente Southern California.

To avoid unnecessary testing, the following were excluded from the patient population: patients with an order or referral for a colonoscopy that was active and authorized within the past 12 months, patients with a future appointment scheduled with a physician in the gastroenterology department, or patients with a colonoscopy completed in the past 5 years. Also excluded were those patients with a terminal illness who were in hospice or palliative care or those in a skilled nursing facility.

Once monthly, the SureNet staff generated from the EMR a list of patients meeting the inclusion and exclusion criteria listed earlier. Those patient charts were then reviewed by a gastroenterologist to determine if a colonoscopy was recommended for diagnosis. Each chart was reviewed to see if 1) there was another documented reason for bleeding (eg, definitive evidence of hemorrhoidal bleeding or anal fissure/ulcer), 2) there were substantial comorbidities that would preclude the patient undergoing colonoscopy, or 3) the patient had seen a gastroenterologist to work up the bleeding where the authors deferred to the original gastroenterologists’ assessment. If, based on these points, a colonoscopy was not recommended, the patient was removed from the outreach list. If a colonoscopy was recommended, the SureNet nurse initiated a telephone encounter to the patient and, using an approved telephone script, informed the patient that because of their rectal bleeding a colonoscopy was recommended. If the patient refused a colonoscopy, the patient was removed from the outreach. If the patient agreed to a colonoscopy, the SureNet nurse initiated a colonoscopy referral, to be signed by the primary care physician. If the SureNet nurse was unable to contact the patient after 3 call attempts, a letter and/or an email were sent to inform the patient of the need to schedule a colonoscopy. For those patients who agreed to a colonoscopy but who had not completed it after 3 months, a letter and/or email was sent to remind the patient to schedule the colonoscopy.

Additional information on patient demographics and patient comorbidities was extracted from the EMR. Statistical analysis was performed using Excel. Relative risks and 95% confidence intervals were estimated to compare different groups.

Results

From December 2014 to December 2019, 9698 eligible patients meeting the inclusion criteria after application of the exclusion criteria were identified. All these patients underwent a chart review by the gastroenterology team, and 5981 (61.7%) were deemed to need a colonoscopy. Of these 5981 patients, 4098 (68.5%) agreed to and were referred for a colonoscopy, and 1430 (34.9%) had a colonoscopy completed during the study period.

Of the 1430 patients who received a colonoscopy after referral by the safety net program, 70.8% (n = 1012) had never had a previous colonoscopy. Those aged 45 to 50 years old were more likely to have never had a previous colonoscopy compared to those aged 51 to 81 years old (95% vs. 59%); 7.5% were found to have had cancer or advanced adenomas, defined as tubular adenoma greater than or equal to 10 mm in size, adenoma with villous histology or high-grade dysplasia. Figure 1 shows the patient characteristics of this population. Also, among these 1430 patients, 34.3% had polyps identified which warranted colonoscopy surveillance more frequently than the routine screening interval recommended by the US Multi-Society Task Force on Colorectal Cancer. ²² Patient characteristics associated with the finding of polyps can be found in Figure 2. Of the 20 cancers that were found, 18 were colorectal cancers, 1 was breast cancer with metastasis to the colon, and 1 was squamous cell anal cancer. Of the 18 colorectal cancers, 10% were diagnosed at stage 0 disease, 10% stage I disease, 30% stage II disease, 25% stage III disease, and 15% stage IV disease.

Figure 1:

Patient characteristics associated with findings of advanced adenomas or cancer by colonoscopy among patients identified in the SureNet system.

Figure 2:

Patient characteristics associated with finding of polyps by colonoscopy among patients identified in the SureNet system.

In those patients where cancer was diagnosed, 55% were male, 45% were White, 25% were Hispanic, 15% were Asian, 10% were Black, and 5% declined to state their race/ethnicity. The most common body mass index (BMI) represented in the cancer population was the obese population (BMI < or = 30) at 50%, followed by normal weight individuals (BMI 18.5–24.0) at 40%, with 5% underweight (BMI < 18.5), and 5% overweight (BMI 25–29.9).

For the 20 patients who were found with cancer, records were manually reviewed to try to determine why the loop may not have been closed. The authors found that 6/20 (30%) had a colonoscopy referral placed, but they failed to show up to the original referral. Another 6/20 (30%) were not offered a colonoscopy, and there was no mention of previous colonoscopy/sigmoidoscopy in the note. Another 5/20 (25%) were not offered a colonoscopy, but a prior normal colonoscopy/sigmoidoscopy was mentioned in the clinician’s progress note regarding the rectal bleeding in the EMR. Of these 5 people, 4 had a colonoscopy noted that was < 5 years old, and 1 person had a normal sigmoidoscopy done a year prior to their rectal bleeding episode. In 2/20 (10%), an anoscopy was done in clinic, but no referral was placed for colonoscopy, and no prior colonoscopy was mentioned in note. One patient (1/10; 5%) had known metastatic breast cancer, and gastroenterology had made a clinical decision that further workup was unnecessary.

Discussion

In this study, leveraging the EMR in an integrated health care delivery system to screen for rectal bleeding without sufficient follow-up, the authors were able to identify 20 patients with cancer as a cause of their rectal bleeding, 87 patients with advanced adenomas, and 490 patients with adenomas that required a change in the frequency of colon cancer screening in the future. The program was implemented with minimal extra resources, albeit with the advantage of starting with an already robust set of SureNet programs and experience in this area.

Although the costs of this program were not estimated, the authors believe that they are small. Staff was not added to the existing SureNet program that consists of 7 full-time employees overseeing 55 SureNet programs. The costs of the “extra” colonoscopies also were not calculated because they should have occurred regardless. The cost of the gastroenterologist chart review was incurred for those patients identified in the EMR as having rectal bleeding. Future iterations of this project could include chart reviews done by advanced practice practitioners or registered nurses, or refinement of EMR searches using natural language processing or artificial intelligence.

Prior SureNet programs focused on closing the loop on abnormal test results. Generally, most programs that address failure to close the loop have focused on test results and not symptoms or signs. The program described here demonstrates that it is possible to also address specific symptoms that are not followed up. Thus, the authors believe that there may be opportunities to identify additional conditions based on signs or symptoms that appear in discrete data fields in an EMR. The authors have also had previous success on closing the loop based on symptoms in a gross hematuria SureNet program. ²³

A prior prospective observational study showed an incidence of colorectal cancer in 3.4% of those who presented to their primary care physician with a concern of rectal bleeding. ²⁴ It is not known what the screening rate for colorectal cancer was in this group. Because the overall colorectal cancer screening rate in the patient population ages 50 years and older ranged from 79.9% to 82.8%, the expectation may be to see a smaller population of those with rectal bleeding have cancerous findings. The cancer detection rates were consistent with findings from groups that may have similar (although not identical) risks. The cancer detection rate of 1.4% was similar to the 1.28% cancer detection rate in those who were fecal immunochemical test positive, who had a colonoscopy 5 to 10 years prior to their positive fecal immunochemical test. ²⁵ Regardless of the exact rate of finding colon cancer compared to prior studies, the authors believe that the rate of finding cancer and advanced adenomas using SureNet for patients with rectal bleeding is high enough to justify this program.

There are some potential limitations to this study. First, it cannot be known what would have happened to these patients had the program not been implemented. The authors did not feel it ethical to include a comparison group without follow-up in a cohort of patients with rectal bleeding. Having said this, the amount of pathology identified seems to be considerable. The yield of the workups was high enough that it would be hard to justify not working up the rectal bleeding in this group of patients. It was not the intent of this study to prove that one needs to work up rectal bleeding with colonoscopy in patients ages 45 to 80 years.

Second, this program relied on diagnoses of rectal bleeding entered in the medical record. If a patient reported rectal bleeding to their physician, but the physician did not enter it as a diagnosis, the patient was not able to be captured in the EMR search, and the patient was not entered into the safety net. Because the program was designed to prioritize specificity over sensitivity, those patients diagnosed with hemorrhoids were excluded, as there is some evidence that this group had a lower likelihood of cancer. ²⁶

It was not known how many of the identified adenomas would have progressed to cancers and how many cancers would have increased in stage without the SureNet program. It is also unknown whether any of the 490 patients with adenomas requiring increased frequency of follow-up colonoscopy would have progressed to cancer if only screened in the usual 10-year intervals.

Finally, this program exists in an integrated health care delivery system that may differ from many other systems. The EMR system provided access to all clinical data, enabling us to not only identify the patients of interest but also identify if they had follow-up or not. It is important to note that one could also obtain both the rectal bleeding and follow-up information from claims data. Because there is not an emergent rush for the follow-up, the lag created by using claims data would likely not compromise outcomes. Thus, systems that can obtain claims data likely could implement something similar to the SureNet program. Additionally, systems that participate in a health information exchange might have similar access to data and could implement a similar program. It is worth noting that a less integrated system of care has implemented a similar system to close the loop on test abnormalities. ²⁴

The studied health care system already has high colon cancer screening rates, with rates for members age 50 to 75 years varying from 79.9% to 82.8% between 2014 and 2019. In systems with lower colon cancer screening rates, a program similar to SureNet may identify greater numbers of patients in need of follow-up, which in turn may aid the discovery of more cancers and advanced adenomas.

Generally, the SureNet programs are well received by clinicians and patients in the system, as they understand the extra layer of safety they provide. Given the low cost and numbers of cancers and adenomas that were obtained, the authors believe that this program adds value and safety to patient care and should be considered by others.