Key to Reducing Unnecessary AXRs: Individualized Metrics

Diagnosis of constipation is primarily based on history and physical exam –not abdominal xray (AXR). In a recent quality improvement study (G Moriel et al. J Pediatr 2020; 225: 109-116. Reducing Abdominal Radiographs to Diagnose Constipation in the Pediatric Emergency Department), ED physicians were trying to improve adherence to evidence-based guidelines for diagnosis of constipation in otherwise healthy children. In this article, the authors note evidence “has shown abdominal radiographs to be unreliable in establishing an association between clinical symptoms of constipation and fecal load on abdominal radiographs.”

As part of the study, the researchers provided two 20-minute presentations to the pediatric emergency department providers and sent emails to them and to resident housestaff. The email for ED provider’s included the provider’s baseline abdominal radiograph frequency. After study was initiated, a followup email was sent with similar information with key information on the project along with individualized data.

Key findings:

  • After the QI interventions, the total percentage of abdominal radiograph decreased to 18% (from 36% at baseline). This 18% decrease was significant ( P < .001) and sustained over a 12-month follow-up period.
  • The average length of stay was 1.07 hours longer for children who had an abdominal radiograph.
  • Clinically important return visits to the emergency department were uncommon during the postintervention phase (125/1830 [6.8%]), and not associated with whether or not an abdominal radiograph was performed at the initial visit.
  • While the study focused on healthy children, the authors noted that the overall population (6 mo-18 years) experienced a decline in AXR usage, regardless of exclusion criteria. At baseline the rate of AXR was 39.5% (1550/3926) which decreased to 20.7% (478/2311).

One interesting piece of data was showing that this intervention resulted in a sustained reduction for 12 months after the intervention observation period, which mitigates the potential influence of the Hawthorne effect.

My take: In my view, the keys to this intervention was providing individualized metrics as well as having leadership in establishing this project. The individualized metrics help physicians recognize when they are outliers and to motivate them to address this.

Related posts:

Why ImproveCareNow is Needed

A few recent articles make a strong argument for collaborative networks, like ImproveCareNow, to improve data collection to determine the most effective therapies.

  1. Kierkus J, et al. JPGN 2015; 60: 580-85.
  2. Audu GK, et al. JPGN 2015; 60: 586-91
  3. Dotson JL, et al. Inflamm Bowel Dis 2015; 21: 1109-14
  4. Saps M, et al. JPGN 2015; 60: 645-53.

A brief description of each study.

1. This study presented a multi-center randomized open-label trial of 99 pediatric patients with Crohn’s disease (CD) who were administered infliximab (IFX) along with an immunomodulator (azathioprine or methotrexate).  After a 10 week induction, 84 were randomized to either monotherapy for 54 weeks or dual therapy for 26 weeks. The authors did not find significant differences in response between the groups.  However, they reached a conclusion: “Twenty-six weeks likely represent (sic) the safe duration of combined IFX/immunomodulator therapy in our sample of pediatric patients with CD.”

2. The second study described three cases of chronic recurrent multifocal ostesomyelitis (CRMO) associated with inflammatory bowel disease.  They tried to identify all pediatric cases in UK in the last 10 years. (As an aside, I have treated one teenager with CRMO and ulcerative colitis.)

3. The third study is a retrospective single center of 30 patients with pediatric Crohn’s disease (CD) who developed intra-abdominal abscesses (IAA) over a 12-year period.  The authors note that this is “the largest single-center review of children and adolescents with CD and IAA to date.” Yet due to the small sample size, the study provides little guidance on this important medical problem; there were no predictors of successful medical or percutaneous drainage therapy.  In addition, with the increasing use of biologics, the authors note that “the issue of which patients will eventually require surgery is even less clear.” Changes in imaging (eg. MRE) and changes in medical management (eg. more enteral nutrition and less corticosteroids) are not discussed.

4. The fourth study is a comprehensive review of randomized placebo-controlled pharmacological clinical trials in children with functional abdominal pain disorders.  They found “no evidence to support the use of most commonly used drugs in children. Only 7 pharmacological RCTs on AP-FGIDs in children were found. Most studies were single center based and had a small sample size.  The methods and outcomes were heterogeneous…We found a considerable risk of bias in most studies…There is an urgent need for well-designed randomized clinical trials using age-appropriate validated outcome measures.”

Each of these studies makes a compelling argument for collaborative research networks.  The first study had a relatively small number of patients, short follow-up period, lack of blinding, and numerous methodological limitations.  How did the authors determine that 26 weeks was the time to stop dual therapy? Among adults with CD, a well-designed SONIC study (NEJM 2010; 362: 1383) showed the superiority of dual therapy during the study period.  In children, because of concerns about thiopurine safety, the best approach is still unclear. The second study identified only three patients despite examining a large population.  Similarly, the third study describes 30 patients with a common complication of CD but provides little insight.

The fourth study is a cautionary tale illustrating the lack of progress due to the absence of collaborative research.  Reports indicate a high prevalence of functional abdominal pain; one study indicated that abdominal pain affects “38% of school children weekly” (J Pediatr 2009; 154: 322-6).  In fact, studies on the high prevalence of this disorder dates back for 60 years (Apley, 1975; Apley & Hale, 1973; Apley & Naish, 1958). Despite the prevalence of this problem, the data for all of the treatments is poor.  The lack of progress in defining treatments for functional abdominal pain is multifactorial, including the following:

  • Cost: For many of the available treatments, there is not a financial incentive to conduct research.
  • Biomarker: lack of objective markers for improvement
  • Disease Stigma: many people attribute functional disorders as being due solely to psychological factors
  • Physician Champions: in pediatric gastroenterology, it took concerted physician efforts over many years to develop ImproveCareNow.  Similar physician champions would be needed to improve the outcomes for children with functional disorders

Bottomline: While ImproveCareNow has a lot of work ahead including improving data reliability and ascertaining accurate outcome measures, I think the effort is forward-thinking and will make a difference in understanding and treating children with IBD.  ImproveCareNow has more than 600 participating pediatric gastroenterologists and more than 20,000 patients. What I would like to see is a sister network to address the morbidity from functional disorders so that in 60 years (or sooner), we will be better equipped to treat children with abdominal pain that is not due to IBD.

Related blog posts:

Fox Theater

Fox Theater

ImproveCareNow Video

A recent (short ~2:30) ImproveCareNow (ICN) Video explains how ImproveCareNow is a forward-thinking network and how it has the potential to lead to better outcomes for children with inflammatory bowel disease.

If you are part of ICN, this video may help explain to your patients what ICN is all about.

What it takes to adopt good ideas

A New Yorker article by Atul Gawande highlights the importance of spending time and talking about good ideas on a person-to-person basis as the best way to help innovations take hold.  Specific innovations that are discussed include the introduction of anesthesia, the control of germs/Lister’s theories of sepsis, adoption of oral rehydration solutions, and preventing hypothermia in newborns. The link to the article and a brief video review on the Colbert report:

CCFA IBD Update -Conference Notes (part 1)

I wanted to share some notes from the Atlanta CCFA “Update in Inflammatory Bowel Disease” conference which was held March 23, 2013.

The first talk by Gil Melmed reviewed immunization issues in IBD.  One of the best quotes of the conference was on his first slide:  “The single most important thing a [rheumatologist] can do to optimize care for a [lupus] patient is to act as their internist, yet that is the last thing most [rheumatologists] wish to or feel qualified to do.” (Wallace DJ. Lupus 2008 (17) 91-2.

While trying to manage all aspects of the health care of our patients with chronic disease is an ideal, he also noted that 20-30% of adult IBD doctors do not understand the importance of avoiding live-virus vaccines (a list of these are noted in previous blog: Protecting the most vulnerable | gutsandgrowth).

One of the take-home points from this talk for me was to avoid use of rotavirus vaccine in newborn patients if their mother is receiving Infliximab or Adalimumab.  Another useful point would be to remind family members of immunosuppressed IBD patients to receive the influenza vaccine rather than the live-virus vaccine (flumist).

The second talk by Wallace Crandall focused on improving health care delivery.  The best quote of the day was from his first slide: “Every system is perfectly designed to achieve exactly the results it gets.”  (Donald Berwick).  One of his slides that I found intriguing showed 2007 data on immunomodulator use within the first 3 months of diagnosis from 10 different centers.  The rate varied from 30% to 98%.  He didn’t try to answer how frequent immunomodulator use in this time period should occur but there was no valid reason for this degree of variation.

In Columbus (Ohio), he stated that they are working to transform chronic care from 15-30 minute visits every 3-6 months to a more continuous process.  “Patients receive only 60% of recommended care.” With a better health care delivery system, this could be improved and potentially improve outcomes.

Another obstacle is the lack of evidence to guide decisions.  Recent “ECCO” guidelines for pediatric ulcerative colitis had 48% of recommendations based on level D evidence.

The third talk by Douglas Drossman was probably the most helpful.  He discussed pain management in IBD.  He reviewed a lot of recent data on irritable bowel syndrome (IBS) and discussed similarities between post-infectious IBS and many patients with treated IBD.

He also discussed central pain aspects and alluded to changes in the brain that happen with chronic treatment (Pain changes brain | gutsandgrowth).  He noted that antidepressants have the potential to reverse brain changes due to chronic pain but that regrowth of neurons can take a long time.  As such, he noted that agents like imipramine are needed for at least one year. Use of imipramine, he noted, has been shown to improve cognitive function in mice with traumatic brain injury (J Neurotrauma 2011; 28: 995).

With regard to IBD, Dr. Drossman noted a wide variability in pain response among patients.  Some who have very significant mucosal disease do not experience significant pain, likely due to downregulation of some central pathways.  He outlined his treatment model for pain in IBD.

  1. Effective communication improves clinical outcomes
  2. Just as there is a “treatment pyramid” for IBD, with IBS the treatment pyramid ncludes the following: motility agents, antispasmodics, antidepressants, psychological intervention, and central augmentation.
  3. With regard to antidepressants, he highlighted the differences between tricyclics, SSRIs, and SNRIs. He noted that SSRIs like citalopram, escitalopram, fluoxetine, paroxetine, and sertraline are not effective at reducing pain but can help anxiety/depression.
  4. Tricyclic antidepressants are effective at reducing pain and are inexpensive.  Side effects can include sedation, dry mouth, dizziness, and constipation.  Desipramine and nortriptyline have fewer side effects.
  5. SNRIs including duloxetine, venlafaxine, desvenlafaxine, and minacipran are effective at reducing pain and fewer side effects.  However, they are considerably more expensive.  Nausea, though, is not infrequent.
  6. Psychologic treatments: cognitive behavioral, psychotherapy, hypnosis, and relaxation training.  When to refer? moderate-severe symptoms, maladaptive coping (eg. “catastrophizing”), and motivated patient.
  7. What is central augmentation? If an antidepressant is not effective, augmentation is adding an agent that may target a different brain receptor.  Potential augmenters include pharmacologic and non-pharmacologic approaches.  Pharmacologics could include added gabapentin, added atypical antipsychotic (e.g.. quetiapine [Ability]), added clonidine, added buspirone, or added mirtazepine (e.g. remeron).  Most practitioners will need the help of a psychiatrist to manage these medications.

Build the information medical highway and expect more traffic

A recent article indicates that increased patient access to online records was associated with increased in-person and telephone contacts (JAMA 2012; 308: 2012-19).

Background (from study introduction): The Institute of Medicine’s report on “Crossing the Quality Chasm” indicated that electronic patient-physician messaging was a promising technology to improve quality and efficiency.  Furthermore, previous studies have suggested that 25-70% of all visits to physicians do not require face-to-face appointments.

Design: To explore this topic further, the authors performed a retrospective cohort study on the use of health care services between 2005-2010 at Kaiser Permanente in Colorado.  This study examined patients ≥18 years old and looked at health care utilization before and after initiation of MyHealthManager (MHM). Users of MHM were compared with nonusers.  And, both groups (users and nonusers) were examined with regard to their health care utilization before and after MHM rollout. The first 30 days before and after activation of MHM were excluded from analysis to minimize the effect of increased utilization at the initiation of MHM.


  • By June 2009, patient use of MHM had increased to 53.8%.  In total, 87,206 MHM patients were identified and 71,663 nonusers were identified for study participation. 
  • MHM users were slightly older and more likely to be female.
  • After initiation of MHM, the rate of office visits increased by 0.7 per member per year (p<0.001) and the number of telephone encounters increased by 0.3 per member per year (p<0.001).  
  • The authors breakdown this data based on age, absence of chronic disease, presence of specific diseases (diabetes, coronary artery disease, congestive heart failure).  In all of these scenarios, MHM users had increased visits after initiation of MHM.  Nonusers generally had the same or less visits at the same time.
  • Figure 2 shows that MHM and nonusers had identical health care utilization beforehand.  Afterwards, the MHM users maintained a parallel line of increased usage that was fairly consistent for a year after rollout.

Why did this happen?  The authors note that the result was contrary to their expectations.  They speculate that individuals may have increased their in-person use after developing additional concerns following their review of information online and that individuals may sign up who are already more likely to use services.  Online access, in these individuals, may facilitate access to more frequent visits.

I think this article points to a more pervasive miscalculation of the effect of information technology and health care utilization.  While electronic health records (EHRs) can help organize and communicate vast amounts of information, the proposition that they will ultimately reduce health care costs/utilization or improve efficiency is looking dubious.  In my opinion, the best we can hope for is that EHRs, when used optimally, will improve the quality of the care.  It is equally possible, however, that EHRs could result in more legible but less accurate information due to well-recognized issues like copy-forwarding with inadequate editing.

Related blog posts:

Quality improvement and better outcomes in Pediatric Inflammatory Bowel Disease

More information on outcomes and quality improvement (QI) efforts are available from Cincinnati Children’s Hospital (JPGN 2012; 55: 679-88).  The QI efforts at Cincinnati are part of a broader effort in QI in pediatric IBD that has been discussed on a previous post (see below).

Our group has rejoined Improve Care Now (ICN) and I look forward to gaining more first hand experience.  Some of the ICN target goals:

  • 80% remission rate among each center and sustained remission rate of 45%
  • 76% steroid-free remission
  • 90% checking TPMT if using thiopurine
  • 95% checking PPD/chest xray if using biologic therapy
  • 95% using accepted methotrexate dosing

As I look at these goals, I wonder whether the remission rate is feasible and about the variation in different centers. For example, among the groups with >75% enrollment of their patients, one center reports a 89% remission rate and another center 64% remission rate.  Given the limited number of therapeutic agents, how could there be such a difference?  Some explanations could include variation in the use of more potent biologic agents as well as capturing/assuring followup of more patients who are doing well.

Given this backdrop, I looked at this most recent publication to learn how their QI practices could translate into better care for my patients.

In this retrospective chart review study of 505 patients, who were followed from 2007-2010 at the IBD center, the remission rate increased from 59% to 76%.  The data were captured prospectively.  This corresponded to improved patient global assessment (>7) from 69% to 80%.  Repeated steroid use dropped from 17% to 10%.  Vitamin D (25-OH D) improved and this also correlated with quiescent disease.

Remission rates were defined by a pediatric Crohn clinical disease index, the sPCDAI, or PUCAI for Ulcerative colitis.

Key observations:

  • There was a trend towards increased use of anti-TNFα therapy (along with decreased use of 6-mercaptopurine) during the study period, but this did not reach statistical significance. No statistical difference was identified in the use of any major IBD medication class.
  • Despite target goals, the only drug class with a statistically significant change in dosing was 5-ASA (from 42 mg/kg to 50 mg/kg).
  • Fecal calprotectin monitoring increased.  The QI team recommended that IBD patients with ‘inactive’ disease have a fecal calprotectin measured every 6 months.  Those with elevated values had therapeutic drug monitoring implemented.  “Fecal calprotectin >400 μg/g is associated with a higher chance of relapse in a pediatric cohort.”
  • Starting in 2009, increased vitamin D monitoring was implemented (every 6 months).  In patients with a serum 25-OHD level < 30 ng/mL, treatment with 50,000 units weekly (for 6-8 weeks) was recommended (8000 units if weight <20 kg).  Once serum 25-OHD was >30 ng/mL, patients were maintained on monthly dosing.
  • Preclinic planning also increased and corresponded to improved remission rates which were 67% in 2009 and 76% in 2010.
  • The authors conclude: “Our results show that significant improvement in patient outcomes were achieved following QI efforts that did not rely on new medications or therapies, rather through initiating novel care processes and standardization of care.”

I think this conclusion is misleading.  First of all, as the authors point out, they did change their therapeutic approach.  They started vitamin D in more patients, used anti-TNFα therapy in more patients, and increased dosing of 5-ASA agents.  Other gains in remission rate could have been related to improved followup of patients who were doing well.

Despite my skepticism about the conclusion, I think the overall achievements are laudable.  Decreasing variation of care and assuring that all patients receive the best care possible with our diagnostic tests and current therapies is certainly worthwhile.  When patient care is studied carefully, this makes sure that “standard” practice like checking TPMT status before thiopurine use, checking PPDs before anti-TNFα therapy, and using optimal drug dosing occur in virtually all patients.

Developing a checklist for each patient can help assure that best practices occur.  For those with electronic medical records, this may mean putting in more “hard stops.”  A hard stop means the physician cannot sign out of a chart without paying attention to a specific detail.  For example, if a physician orders methotrexate, a “hard stop” could come up if the dosing was not in the typical target range.

Useful link:

Link for disease classification (quiescent, mild, moderate, severe):

Related blog entries:

Improve Care Now

Several years ago, “ImproveCareNow” Network was established with a goal of improving the quality of care children with inflammatory bowel disease. (

Documentation of some outcomes of this effort have now been published (Pediatrics 2012; 129: e1030-e1041).  In this study from six U.S. centers, data from 843 children with Crohn’s disease and Ulcerative Colitis were available.  Centers were eligible to participate if they did not have extensive quality improvement (QI) experience and if they were able to enroll >75% of their IBD patients.  During the course of the study (2007-2010), there were increases in the proportion of patients with measurement of thiopurine methyltransferase (TPMT) before initiation of thiopurines, increased percentage of individuals receiving an appropriate thiopurine dosage, and an increased percentage of patients with inactive disease.  In Crohn’s disease, the number with inactive disease changed from 55% to 68%; in ulcerative colitis, the number increased from 61% to 72%.  

This study shows significant progress in these patients and the potential benefit of a more-structured approach.  However, some of the improvements in these outcomes may not be to process improvements. 

Study limitations:

  • The study does not document the rate of inactive disease in non-participating centers; thus there is not an adequate control population.  Given more widespread use of biologic therapies, this may account for much of the improvement in outcomes.   In fact, the study does not describe the percentage of patients receiving thiopurines or biologic agents.
  • The use of the physician global assessment (PGA) as indicative of disease activity.  The authors acknowledge that “PGA ….is a relatively subjective measure.”  A more reliable measure of disease activity like a stool calprotectin, endoscopy, or imaging study improvement would have been helpful.  
  • Better documentation or better outcomes? It is not clear whether some of the improvement reflects better documentation or improved care delivery.

Despite the limitations of this study, it is a good starting point.  The goal of improving the care of patients is shared by almost all physicians.  One way to start making a difference is measuring specific outcomes and targeting specific goals/checklists. Going forward with QI there are many QI hurdles

  • Defining optimal care is nearly impossible, there are few trials comparing efficacy & difficult to judge risk/benefit ratio of many therapies.
  • Difficulty selecting appropriate outcome measures -indications for hospitalization and surgery are not clear-cut in many cases.  Surgery may be an appropriate treatment rather than a negative outcome. 
  • Risk adjustment is an imperfect science.

Goals with Quality Improvement:

  • Plan-Do-Study-Act cycles.
  • Learning from high-performing centers -benchmarking

Warranted vs unwarranted variations:
Unwarranted: selecting inadequate dosing of medications, failure to check for TB before remicade/biologics, failure to consider drug interactions, failure to discuss options with families
Warranted: variation due to differences in disease (eg budesonide for ileitis but not pancolitis), variation driven by patient preferences (eg. using NG instead of corticosteroids)

“When you can measure what you are speaking about and express it in numbers, you know something about it; but when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind” –Lord Kelvin 1883

“Not everything that counts can be counted, and not everything that can be counted counts.” –Albert Einstein

Additional references on quality improvement:

  • Free Self Management Handbook endorsed by ImproveCareNow:

  • -JPGN 2009; 49: 272.  Review of quality measures/history.
  • -Kohn LT, et al. To Err is Human: Building Safer Health System.  Nat’l Academy Press; 2000
  • -Crossing the quality chasm.  Wash DC: Nat’l Academy Press; 2001.
  • -Clin Gastro & Hep 2010; 8: 709.  Quality indicator sets in cirrhosis.
  • -J Pediatr 2005; 146: 744-50.  Bucuvalas JC et al.  Adjusting calcinerin inhibitor dosing
  • -IBD guidelines:
  • -JPGN 2009; 49: 297.  Variation in care in pediatric Crohn disease. (Our center participated in this study)
  • -J Pediatr 2009; 154: 582.  Ventilator assoc pneumonia reduced w bundle of hand hygiene, elevated HOB, not changing ventilator circuits except when soiled & mouth care
  • -“The Learning Curve”(Atul Gawande) in The Best American Essays 2003 -some CF ctrs outperform, extending life expectancy ~14yrs.
  • -” A Surgeons Notes on Performance-Atul Gawande
    -NEJM 2007; 357: 2652.  Nice editorial.

Comments from lead author forwarded to blog:

Thanks for including our recent paper in your blog.  I agree with the study limitations as you mentioned, but wanted to put a couple of things in context.  We actually thought very hard about the issue of a control group, and specifically how to determine what the remission rate would be without the intervention.  We could not come up with a reasonable comparator (very few people/groups outside of ICN even know what the remission rate of their population is).  One very rough measure (which we did not feel was legitimate to include) is the rate of remission for new sites as they are joining the collaborative.  Established sites with complete participation have better outcomes than newer sites or sites that participate less fully, suggesting that at least a significant portion of the effect is due to the QI intervention rather than secular trend.  As far as the objective measures for outcomes, I know you realize that we are not able to mandate specific objective evaluations for a QI/observational research study, so we may never be able to use mucosal healing (for example) as our measure of remission.  However you probably noticed in the paper that we did demonstrate improvements in objective measures including sPCDAI, PUCAI and use of corticosteroids.  Certainly improvements in process measures may have been due in part to better documentation, but it is unlikely that objective outcome measures improved due to documentation. Thanks again for sharing our work with your audience.