Charna Albert
November 2024—In health care, information overload is hardly a novel problem. But when it comes to overlooking impaired kidney function in the electronic health record, Children’s Mercy Kansas City found a novel fix—at least for a pediatric institution.
The solution, says Darcy Weidemann, MD, MHS, pediatric nephrologist at Children’s Mercy Kansas City and associate professor, University of Missouri-Kansas City School of Medicine, was to implement estimated glomerular filtration rate in the EHR.
“As far as I’m aware we’re the first pediatric institution to have it directly integrated,” says Dr. Weidemann, who spoke about her institution’s experience at the ADLM meeting in July and in an interview with CAP TODAY. “I have to imagine it’s coming because of the benefit, but it’s an incredible amount of work to make it happen.” (Dr. Weidemann co-presented with Angela Ferguson, PhD, D[ABCC], also of Children’s Mercy. For how the laboratory made it happen, see https://www.captodayonline.com/egfr-equations-in-the-ehr-how-lab-met-the-request.)
Yet the work didn’t end there for Dr. Weidemann and her colleagues in nephrology and the laboratory. Adding eGFR to the system presented new opportunities to build clinical decision support, says Dr. Weidemann, who is also co-director of the Quality, Safety, and Operational Excellence Improvement Program at Children’s Mercy Kansas City.
The new projects—beginning with the eGFR implementation in 2022—were set in motion after several adverse events related to impaired kidney function occurred in the hospital. Patients whose kidney disease had gone unrecognized were given potassium-containing IV fluids without laboratory monitoring, vancomycin was administered without renal dosing, and multiple nephrotoxic medications were prescribed without appropriate laboratory monitoring.
“As we were exploring these events,” Dr. Weidemann says, “we realized one of the big problems was that people did not realize these patients had impaired kidney function.” In the chronic kidney disease population, “we make sure the problem lists are updated,” she says. “But you can get lost in the signal-to-noise ratio sometimes, with all your tasks at hand.”
Another source of confusion: the pediatric reference ranges for serum creatinine. “Even though you have age-based cutoffs, this is not data that’s generated on thousands of kids,” she says. “You can often miss mild chronic kidney disease, where the GFR is in that 60 to 90 range, because there’s so much variation in body habitus.” For a short two-year-old, for example, a creatinine of 0.6 mg/dL would be too high. “If it doesn’t flag, for somebody who doesn’t live and think and dream about creatinine all day, that’s an easy thing to overlook.”
Lack of education, too, contributes. “Our medical students only get eight weeks of pediatrics. We’re lucky if they get any exposure to our patients in their medical training. And the intricacies of GFR and how we estimate GFR are very different from how they are in the adult world.” The Accreditation Council for Graduate Medical Education is changing its licensing rules, too, leaving residents with less inpatient exposure. “Starting in July 2026 we won’t have residents rotating with us. So I think we’re going to face additional barriers and challenges moving forward with appropriate exposure to these patients.”
Dr. Weidemann opened with the case of a 12-year-old girl with spina bifida who presented in the emergency room with fever, emesis, and flank pain. The patient had a history of recurrent urinary tract infections and a neurogenic bladder and was on sulfamethoxazole-trimethoprim for UTI prophylaxis.
Her urinalysis was suggestive of a severe urinary tract infection, with a 3+ leukocyte esterase. Microscopy showed greater than 100 red and 100 white blood cells per high power field. A renal ultrasound revealed a left lower pole lesion or mass and echogenic renal parenchyma with poor corticomedullary differentiation. The radiologist recommended further visualization with a CT or MRI.
“This patient has decreased muscle mass, she’s wheelchair bound, she’s nonmobile,” Dr. Weidemann says. Her creatinine of 0.5 mg/dL did not flag as abnormal, and her eGFR-creatinine was close to normal, at 90.2 mL/min/1.73 m2. Her cystatin C, however, was 2.1 mg/L. “I was able to recommend cystatin C, and we are fortunate because our lab is quick—I can usually get cystatin C within 60 to 90 minutes.” Her eGFR-cystatin was 38 mL/min/1.73 m2, making her average eGFR 64.1 mL/min/1.73 m2. “And this was incredibly useful information. Even overnight we were able to figure out a plan.”
They decided to defer the MRI, given the patient’s low GFR and concern about exposure to contrast, and lowered her prophylaxis dose of Bactrim. The patient had substantially impaired kidney function that had gone unrecognized, Dr. Weidemann says. “She had never seen a nephrologist. She never knew she had kidney disease. And here she is pushing stage four almost.”
Substantial evidence shows that integrating eGFR into the medical record can improve recognition of impaired kidney function, Dr. Weidemann says, citing a study that found that presenting physicians treating children with calculated eGFR values in addition to serum creatinine more than quadrupled the number who accurately estimated kidney function (Schnapp A, et al. Pediatr Nephrol. 2023;38[9]:3091–3097). Thirty-nine percent of physicians (board-certified in pediatrics) correctly assessed kidney function when presented with eGFR values but misinterpreted it when equivalent creatinine values were presented. “That is certainly statistically significant,” she says, “but it’s also clinically significant.”
In addition to adding eGFR into the EHR, Dr. Weidemann and her colleagues aimed to increase the number of cystatin C orders and to develop new clinical decision support for the use of contrast agents and for orders of nephrotoxic medication. But before they could begin their work, they had to ensure patient height was being captured accurately in the EHR. Dr. Weidemann and her colleagues updated all 190 of the hospital’s admission PowerPlans in the Cerner EHR to have the height or length default checked upon admission. They also standardized nursing and medical assistant education and created job aids to assist with proper height and length assessment, including the two-person technique for infants and segmental length assessment for nonambulatory patients.
Dr. Weidemann shared the case of a two-year-old girl with steroid-resistant nephrotic syndrome. She was discharged from the hospital after a three-week stay in which she received steroids and multiple IV albumin and furosemide infusions to control her edema. She was also started on tacrolimus to try to induce remission.
When Dr. Weidemann saw her for routine follow-up three weeks post-discharge, the patient’s creatinine was 0.41 mg/dL. Her previous creatinine was 0.26 mg/dL. “She had a 58 percent increase in her serum creatinine,” Dr. Weidemann says. But that change was made much more obvious by her eGFR-creatinine, which was 68 mL/min/1.73 m2. “And that’s a big deal in a patient like this, who has a five-year risk of end-stage kidney disease and being on dialysis. Even with the best treatment she has a challenging disease process to manage.”
The patient’s potassium had increased from 4.4 to 5.3. Her tacrolimus level, too, was elevated. “We’re starting to see a little bit of tacrolimus toxicity and that’s probably why her potassium was elevated.”
The American College of Radiology and National Kidney Foundation released in 2020 and 2021 consensus statements on the use of iodinated and gadolinium-based contrast in patients with kidney disease (Davenport MS, et al. Radiology. 2020;294[3]:660–668; Weinreb JC, et al. Radiology. 2021;298[1]:28–35).
According to the 2021 statement, the risk of nephrogenic systemic fibrosis from group two gadolinium-based contrast media in patients with advanced kidney disease is thought to be low.
Depending on the clinical indication, it says, the potential harms of delaying or withholding group two or group three gadolinium-based contrast media for an MRI in a patient with an acute kidney injury or eGFR less than 30 mL/min/1.73 m2 should be balanced against and may outweigh the risk of nephrogenic systemic fibrosis.
“The guidelines recommend the use of class two agents and that you don’t even have to check kidney function anymore,” Dr. Weidemann says. “We haven’t quite gotten to that level of ease. We recommend the use of class three agents and to still check kidney function, just because there’s less safety data” on patients with eGFR under 30 mL/min/1.73 m2. “Regardless, we decided to incorporate many of these evidence-based guidelines into our care process models.” The 2020 consensus statement, she says, notes that prophylaxis with intravenous normal saline is indicated for patients who have acute kidney injury or an eGFR less than 30 mL/ min/1.73 m2.
The new safety alert and care process model Dr. Weidemann and her colleagues developed works as follows: When an MRI is ordered, the EHR prompts the tech who orders the contrast to check if kidney impairment screening is indicated. If the patient has a history of acute kidney injury, CKD, or kidney surgery or ablation, or if the patient is at risk for acute kidney injury, the tech is prompted to order an eGFR-creatinine. The care process model includes a link to an online eGFR calculator for outside providers who may not use the same EHR the hospital uses, Dr. Weidemann says, and therefore can’t order calculated eGFR.
If the patient’s eGFR is greater than 30 mL/min/1.73 m2, the patient can complete the MRI. If it’s less than 30 mL/min/1.73 m2, the care team discusses the case with nephrology before deciding whether to proceed. They developed a similar model for CT scans, which also includes recommendations on pre- and post-hydration. When a CT scan is ordered, the system provides a link to a Cerner PowerPlan with the pre-/post-iodinated contrast hydration protocol.
As Dr. Weidemann and her colleagues worked on a new clinical decision support tool for medication orders, they discovered that the Cerner application for pharmacy verification provided pharmacists with patient eGFR using the Schwartz “bedside” equation. “The pharmacists always had the eGFR on the back end and the clinicians never had it,” Dr. Weidemann says. “And as we learned, it wasn’t available to clinicians during their ordering.”
They decided to use abnormal creatinine as the trigger for the new alert, rather than eGFR or cystatin C. “Eighty percent of people admitted to the hospital are going to be prescribed at least one nephrotoxic medication,” Dr. Weidemann says. “And we do not have GFR estimating equations developed in healthy children, so there was a lot of conversation about whether or not they’re valid in patients who don’t have kidney disease.” (The CKiD U25 equation is less precise in patients with an eGFR greater than 75 mL/min/1.73 m2.)
Many of the drug dosing guidelines, too, are based on older pharmacokinetic studies and previous GFR-estimating equations. “The data providing normal values in children is not great, and a lot of the data for drug safety and proper dosing and kidney function is even worse than the normal reference ranges. And a lot of the foundational work with these medications is based on the Cockcroft-Gault estimating equations,” she says. “So it was apples to oranges.”
Ultimately, they decided abnormal creatinine would be sufficient. “If your GFR is less than 50 or 60, that’s typically when medical adjustments need to happen,” she says. Abnormal creatinine captures more than 95 percent of clinical scenarios with eGFR under 60, she says, which is what matters most in medication dosing.
“It was a compromise to get something functional in the system,” she adds.
The order alert works as follows: If a nephrotoxic medication such as vancomycin is ordered and the patient’s creatinine is outside the normal reference range, an alert is triggered to notify the ordering physician. “And then there are links to cancel the vancomycin, you can override the alert, you can modify it, or you can automatically place an order for eGFR-creatinine.”
“Our first week was kind of a bust,” she says. “We started with Cerner’s list of 309 renal drugs provided to us by the EMR, and a lot of the ones that were alerting weren’t useful.” They worked with nephrology and the medication safety officer to refine the list, narrowing it to 120 drugs. Many of the chemotherapy drugs for patients with hematological cancers have a separate verification process, they learned, with chemotherapy dosing occurring far in advance. “We ended up excluding all the chemotherapy PowerPlans,” she says.
They now receive about 1,000 alerts per month. In 90.5 percent, the ordering physician overrides the alert and continues the order, “which I think is probably appropriate,” she says. The order is canceled 5.8 percent of the time, and 3.6 percent of orders are modified. “But what I’m most impressed with is that 53.3 percent justify their decision, if they continue, with a voluntary free-text response.” Most of these responses note that the orderer has verified with the pharmacist that the medication is safe.
“I would love to get to a point where it was a smarter system that was able to integrate the known GFR dosing,” Dr. Weidemann says. “Let’s say, for example, a patient has a GFR of 34,” and a physician is attempting to prescribe a nephrotoxic medication. The computer would then say, “‘You’re dosing this medication at 50 per kilo, and it needs to be 25 per kilo’—to have the correct dose based on eGFR in there and provide that more personalized education on what is correct. Right now, it’s just an alert that says, ‘Please double-check through your regular avenues.’”
A good GFR estimating equation for healthy children would improve recognition of impaired kidney function, Dr. Weidemann notes, but expanding the research on kidney disease in children to a healthy population would be difficult. “It’s been millions of dollars of effort from NIH to fund this, to have decent estimating equations in children,” she says. “Measured GFR is incredibly time-consuming and expensive. You have to get an IV, you have to do plasma disappearance of iohexol. So it’s asking somebody to commit to a six-hour hospital stay. Who’s going to be able to fund that, or agree to that?”
For Dr. Weidemann and her colleagues in nephrology, integrating outside labs into the EHR is now a focus.
“We have a huge referral area, and we see patients who travel six, eight hours to get to us. It’s not uncommon for patients to get their labs done locally and then they get sent to us for review. So we’re working on a standardized hospital process to deal with these outside labs,” she says. Previously, the process required scanning the outside labs into the hospital system. “But then you had to go to this other program to review it, and it was just an image, so being able to look at lab trends was virtually impossible,” she says.
The solution: Nephrology administrative staff now input all outside labs into Cerner. “When we go to the lab tab on our results review, there’s a section within that lab area that has all the previous labs. And you can graph that data and look at trends over time,” she says. “It’s a lot of administrative work, but for us it has improved care.”
Some of the other hospital divisions have followed suit, she says, entering outside laboratory data in the system. “There has been some trickle-down effect to other groups. But every division has its own resources, so it’s unfortunately not yet standardized across the hospital.”
Where they have seen uptake outside their division is in awareness of eGFR. “Our next step is partnering with other sections,” she says, particularly those that care for high-risk populations, such as the urology and spinal differences clinic, the nephrology-oncology multidisciplinary survivor’s clinic, and solid organ transplant. Patient education initiatives are another focus. They’re working on implementing a critical action note in the EHR with the last eGFR updated and automatic linkage for patient education materials by ICD-10 codes. And they hope to distribute wallet cards for patients that would include eGFR, for emergency room use.
In late August they began calculating eGFR from outside laboratory data. If an outside laboratory orders a cystatin C, eGFR-cystatin will automatically be calculated for the physician. As with internal ordering, eGFR-creatinine is not calculated automatically, as Dr. Weidemann’s co-presenter explained. One complication is that height has to be current. “So there’s still a gap,” Dr. Weidemann says, for patients who live outside the immediate area and don’t have a recent height in the system.
Dr. Weidemann returned to the case of the patient with spina bifida and the presumed kidney mass. “We had a long discussion among all the multidisciplinary teams about whether to get an MRI versus a CT scan,” eventually settling on the latter. The presumed mass, the scan revealed, was a lobulated kidney with multiple cortical scars from previous recurrent episodes of pyelonephritis. “So she did not have a mass, which was a big relief to the family. But she has chronic kidney disease, she’s now progressed to stage four, and she was just listed for a kidney transplant.”
The collaboration with the laboratory to implement eGFR in the medical record is what Dr. Weidemann credits as the jumping-off point for the new clinical decision support.
“It spurred a number of high-impact clinical interventions that continue to this day.”
Charna Albert is CAP TODAY associate contributing editor.