Monday, November 22, 2021

Race-based clinical prediction tools are ripe for reassessment

In June, Dr. Bonzo Reddick's editorial "Fallacies and Dangers of Practicing Race-Based Medicine" reviewed the limitations of several commonly used clinical prediction tools that employ race as a biologic variable rather than recognizing it as a social construct. For example, he pointed out that the American College of Cardiology / American Heart Association Pooled Cohort Equations (PCE) predict that "a 40-year-old White male smoker has a lower cardiovascular risk than a 40-year-old Black male nonsmoker," or put more bluntly, "being a Black man is more dangerous than smoking." Since then, researchers and policy makers have made considerable progress in addressing the inappropriate use of race in medical decision making.

A Curbside Consultation in the September issue of American Family Physician introduced a multiracial patient who is confused by the need to identify as African American, White, or Other so that his clinician can evaluate the appropriateness of statin therapy. If White, his estimated 10-year cardiovascular disease (CVD) risk would be 5.8%; if African American, it would be 17.7%. Similarly, a preprint study using thousands of hypothetical and actual patients concluded that large differences in PCE estimates in Black versus White persons with identical risk factor profiles would have the practical effect of "introduc[ing] race-related variations in clinical recommendations for CVD prevention." Until new cardiovascular risk prediction models are developed that omit race, Drs. Mara Gordon and Isha Marina Di Bartolo suggested that physicians exercise caution when using race as a marker of genetic ancestry; consider alternative approaches to risk stratification; and use social determinants of health as an alternative to demographics.

Turning from the heart to the kidneys, including race in the estimation of glomerular filtration rate (eGFR) has the effect of increasing a Black person's eGFR relative to a White person's with the same serum creatinine level. Consequently, Black patients with chronic kidney disease become eligible for kidney transplants nearly two years later than their White counterparts. Underlying this point, Glenda Roberts, a patient representative to a National Kidney Foundation and American Society of Nephrology Task Force that recommended implementing a refitted eGFR calculation that does not include race, observed in a recent opinion piece that though she self-identifies as Black, learning from a DNA analysis that her ancestry was only 48% African (making her, technically, White) would have gotten her on the transplant list sooner! The Chronic Kidney Disease Epidemiology Collaboration has published new eGFR equations that omit race and incorporate serum creatinine and cystatin C.

A 2007 calculator for predicting the likelihood of a successful vaginal birth after cesarean (VBAC) delivery that includes race-based correction factors for African American and Hispanic women was later challenged for promoting disparities in cesarean rates. Earlier this year, researchers from the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network unveiled a new VBAC calculator without race variables that has excellent calibration and a similar area under the receiver operating characteristic curve as the previous calculator.

Further work remains to be done. The Agency for Healthcare Research and Quality (AHRQ) posted draft Key Questions for a future systematic evidence report on the impact of clinical algorithms on racial disparities in health and health care. Another AHRQ-funded methods report on racism and health inequities in clinical preventive services and guideline development supported the U.S. Preventive Services Task Force's proposed changes to its recommendation processes to mitigate the effects of systemic racism.

**

This post first appeared on the AFP Community Blog.

Monday, November 15, 2021

Neonatal abstinence syndrome: another consequence of the opioid epidemic

The opioid epidemic has accelerated during the COVID-19 pandemic, with the Centers for Disease Control and Prevention estimating that more than 93,000 people died from opioid-related overdoses in 2020, a staggering 30 percent increase over 2019. As more pregnant patients have been using opioids, rates of neonatal abstinence syndrome (NAS) have also been on the rise, nearly doubling between 2010 and 2017. Family physicians who care for newborns will increasingly be called on to manage this syndrome. In an editorial in the September issue of American Family Physician, Drs. Roschanak Mossabeb and Kevin Sowti reviewed key points in treatment of NAS, including a low-stimulation environment, skin-to-skin contact, frequent breastfeeding, and opioid therapy when indicated. They emphasized that involving the newborn's mother in the care plan is essential to achieving the best outcomes:

Mothers should be viewed as medicine for their infants; by spending time together, infants will likely need less pharmacologic treatment, hence a shorter hospital stay and decreased hospital costs. In addition, strengthening the mother-infant bond may reduce postpartum depression and improve maternal stress response.

Unfortunately, for a variety of reasons mothers and newborns with NAS are often separated after being discharged home. A recent county-level analysis in Health Affairs found that national increases in NAS are associated with increases in placement of infants in foster care: "every one diagnosis ... per ten births was associated with a 41 percent higher rate of infant foster care entry." Infants residing in rural counties were more likely to be placed in foster care than those residing in urban counties.

A quality improvement collaborative in Colorado hospitals aimed to standardize care of opioid-exposed newborns by implementing the Eat, Sleep, Console model. Study results showed that while hospital length of stay and pharmacologic therapy use decreased for all mother-infant dyads during the study period, these positive effects were delayed in mothers who self-identified as being of Hispanic ethnicity. Racial differences in treatment and outcomes have also been observed in studies comparing Black and and White newborns with NAS. However, a critique of two of these studies cautioned physicians against conflating racial disparities with genetic differences in treatment requirements, noting that implicit bias is more likely to explain the disparities than biologically-based explanations.

**

This post first appeared on the AFP Community Blog.

Tuesday, November 9, 2021

Virtual-first primary care: will the proof be in the pudding?

Six months ago, Dr. Elisabeth Rosenthal, editor-in-chief of Kaiser Health News and a former emergency medicine physician, wrote a commentary titled "Telemedicine Is a Tool - Not a Replacement for Your Doctor's Touch." She warned that since the expansion of telehealth driven by the exigencies of the COVID-19 pandemic had outstripped the evidence of its effectiveness and cost-effectiveness, "deploying it too widely risks poorer care, inequities and even more outrageous charges in a system already infamous for big bills." At that time, my office was transitioning back to seeing a majority of patients in person, a process that has continued to the point that virtual (video or audio-only) visits now make up only about 10-20% of my daily schedule. I have welcomed this return to normal. Like nearly all clinicians, I experienced my share of awkward telemedicine tales and never quite figured out how to virtually diagnose and manage conditions that require a hands-on physical examination, such as musculoskeletal issues. My telehealth experience also confirmed my prior suspicion that connecting with new patients via tablet or computer screen was more difficult than in the physical proximity of an examination room.

In contrast, some health care insurers and vendors are going all-in on virtual care, offering "virtual-first" plans with lower premiums that steer patients to primary care physicians whom they may never meet in person. Teladoc offers Primary360; Amwell has Virtual Primary Care; CareFirst is advertising CloseKnit; and United Healthcare is launching NavigateNOW. All of these services offer more scheduling flexibility than traditional practices, self-monitoring tools, and promise to refer patients for in person visits with primary care or specialist physicians if necessary. It seems that these plans are targeted at younger adults, who are less likely to have an established source of primary care and more likely to be comfortable having solely digital relationships with physicians. (A research letter published this week in Annals of Internal Medicine found that fewer than 1 in 4 U.S. adults over age 65 used telemedicine services in 2020.)

This phenomenon raises a number of questions. Who are these companies hiring to provide care? I periodically receive unsolicited e-mail requests to moonlight for commercial telehealth providers that tout flexible schedules and generous payment rates, but the idea of sitting in front of a screen any more than I already am makes it a nonstarter. I very much doubt there is an existing pool of doctors or nurse practitioners who have been doing nothing during the pandemic. And although some have suggested that practicing virtual medicine well requires specific training, core competencies and curricula remain to be developed. That means that there's no way to assure that present-day "medical virtualists" are actually well-qualified to practice in this challenging environment.

Drs. Kurt Herzer and Peter Pronovost have argued in a JAMA Viewpoint that like in-person care, it's important to demonstrate that virtual care is high-quality: that is, safe and effective, efficient and timely, patient centered and equitable. None of these are so clearly true that the research doesn't need to be done to find out. It's quite possible, in fact, that virtual-first primary care could lead to more incorrect diagnoses, unnecessary visits, and excessive testing, and worsen health disparities due to inequitable access. 

As a believer in the value of a longitudinal relationship with a primary care clinician, I think that if virtual-first plans end up creating such relationships where there wouldn't have been any, then they will likely improve health for some people. If, on the other hand, these plans lure large numbers of patients away from brick-and-mortar practices due to convenience, costs, or reduced choices (e.g., being the only type of plan offered by an employer), I forecast that accomplishing key primary care tasks such as care coordination will become even more difficult. In the past two decades, patients enrolled in fee-for-service Medicare saw more outpatient specialists without increasing their contacts with primary care physicians. It's hard to imagine that a virtual primary care doctor in California (or India or China, for that matter) is going to have more than a superficial knowledge of subspecialists, behavioral health, or health care resources available in Washington, DC. But like it or not, the future of primary care is coming, and time will tell if it makes things better or worse for us all.

Monday, November 8, 2021

Trust and relationships in primary care

The editor of the Annals of Family Medicine invited me to write an editorial connecting and reflecting on several articles in the journal's November/December 2021 issue. That editorial, "Trust and relationships remain at the heart of primary care," was published online today. I'm proud of what I wrote, and I hope that some of you will take the time to read it and let me know what you think.

Monday, October 25, 2021

Measuring and minimizing low-value health care

Hospitals and health systems have often needed to restrict nonemergent care during COVID-19 surges, with mixed effects on patients. Some patients may experience worse outcomes when necessary treatment or surgery is postponed, while others may avoid receiving unnecessary and potentially harmful (low-value) care. Of course, relying on a pandemic to reduce low-value care is not a strategy; at best, it's a blunt instrument that will be discarded when the public health emergency ends. Recent studies of pre-pandemic low-value care have further demonstrated the need for sustainable interventions.

Dr. Ishani Ganguli and colleagues described the use of 41 low-value medical services in a retrospective cohort of more than 11 million Medicare beneficiaries across 556 health systems. They found that the most common services were preoperative laboratory testing, prostate-specific antigen testing in men older than 70 years, and antipsychotic medications in patients with dementia. Characteristics of health systems associated with greater low-value care (based on a composite measure of the 28 most common services) were having a smaller proportion of primary care physicians, a larger proportion of patients of color, no teaching hospital, higher health care spending, and headquarters in the Southern or Western U.S.

Another recent study of Medicare claims data examined the prevalence and costs of hospital-acquired conditions and patient safety indicator events associated with a selection of low-value inpatient procedures. The investigators identified 231 hospital-acquired conditions and 1,764 patient safety indicator events associated with these procedures from 2016 to 2018, resulting in $3.16 million and $26.7 million in additional health care costs, respectively. For example, hospital-acquired conditions occurring during an admission for percutaneous coronary intervention extended length of stay by an average of 17.5 days and increased the cost of hospitalization by $22,000. Those are impressive amounts given that the procedure may have been unnecessary in the first place!

The Cochrane Library has created a special collection of systematic reviews on resource-intensive interventions "for which there is high or moderate certainty evidence that they confer clinically small or no effects, and for which there is some evidence of harm to patients." Examples include preoperative testing for cataract surgery, percutaneous vertebroplasty for vertebral compression fractures, and intensive follow-up strategies after treatment of non-metastatic colorectal cancer.

As Dr. Jennifer Middleton pointed out in a 2018 editorial, changing physician behavior to minimize low-value care requires creating new workflows and systems of care. A systematic review of 131 articles on Choosing Wisely interventions in the U.S. through June 2019 found that the most effective interventions target clinicians rather than patients, are active rather than passive, and include multiple components. The type of low-value service targeted did not affect outcomes. Components of effective clinician-focused interventions included behavioral nudges, feedback / report cards, clinical decision support, electronic health record enhancements, clinician champions, education and academic detailing, and creating new clinical pathways.

**

This post first appeared on the AFP Community Blog.

Tuesday, October 19, 2021

The past, present and future of cancer screening

I've given a lecture in Georgetown University School of Medicine's Evidence-Based Medicine course on evaluating screening tests every fall for the past decade, and for several years prior to that I taught these principles to public health students at Johns Hopkins University. I link the origin of evidence-based assessment of clinical preventive services to the creation of the U.S. Preventive Services Task Force (USPSTF) in 1984, first under the aegis of the U.S. Public Health Service and later, the Agency for Healthcare Research and Quality. In doing so, I imply that prior to that date, screening in primary care was in the stone ages. As it turns out, that's not exactly true.

The recent digitization of archives of The Journal of Family Practice dating back to 1974 allowed me to discover a 4-part series of previously inaccessible papers by Drs. Paul Frame and Stephen Carlson titled "A Critical Review of Periodic Health Screening Using Specific Screening Criteria." These historical articles, published before I was born, analyzed screening tests for 36 selected diseases using criteria that would be familiar to anyone studying screening today:

1. The disease must have a significant effect on quality or quantity of life.

2. Acceptable methods of treatment must be available.

3. The disease must have an asymptomatic period during which detection and treatment significantly reduce morbidity and/or mortality.

4. Treatment in the asymptomatic phase must yield a therapeutic result superior to that obtained by delaying treatment until symptoms appear.

5. Tests must be available at reasonable cost to detect the condition in the asymptomatic period.

6. The incidence of the condition must be sufficient to justify the cost of screening.

Here are some of their recommendations for screening in adults that remain essentially unchanged half a century later: Take a smoking history. Measure height and weight. Check blood pressure every 2 years. Don't routinely do EKGs in asymptomatic people. Measure cholesterol levels. Don't screen for lung cancer with chest x-ray or sputum cytology (these days we screen high-risk persons with CT). Don't screen for brain tumors. Don't screen for COPD. Don't screen for cirrhosis (advanced liver disease). Don't screen for oral cancer, stomach cancer, or pancreatic cancer. Perform fecal occult blood testing to screen for colorectal cancer (this remains an option although many patients now choose screening colonoscopy). Don't screen for bacteriuria in nonpregnant patients. Screen for syphilis in at-risk persons. Don't screen for testicular cancer, bladder cancer, or kidney cancer. Do Pap smears for cervical cancer in women older than age 20 (today we may add or substitute HPV testing in certain patients). Don't screen for endometrial or ovarian cancer.

In terms of screening for cancer, there were only two substantial changes from these 1975 guidelines to today's accepted practices. For breast cancer, we now recommend biennial mammography for all women starting at age 50 (or age 40 if desired) and discourage routine breast self-examination, which Frame and Carlson endorsed due to it being a benign intervention - it hadn't yet been shown to be harmful in terms of increasing false positive results and biopsies. And while Frame and Carlson discouraged prostate cancer screening, the current USPSTF thinks that selective screening is acceptable in men age 55-69 years (although I disagree).

The fact is, though, most cancers aren't diagnosable or diagnosed through screening. A a recent National Academies workshop, Dr. Chyke Doubeni noted that less than half of all cancer deaths are potentially preventable through current screening techniques. Dr. Elizabeth Sarma's research showed that only one-third of estimated new cancer cases this year are potentially screen detectable. For example, data show that 67 to 82 percent of colorectal cancers are detected after patients present with symptoms, not due to an abnormal screening test result. This imbalance would improve a little if 100% of eligible adults got screened, but the fact is that current technology can't prevent every cancer, or even the majority of them.

That's a quick tour of the past and present. What about the future of cancer screening? There is a lot of optimism these days about multicancer early detection (MCED) tests, also called "liquid biopsies." Imagine this: your doctor would be able to take a single blood sample and test it for multiple cancers simultaneously, including cancers that currently have no effective standard screening test. That sounds great, but there are lots of potential pitfalls, including the inability of some tests to identify cancer tissue of origin (so you could get a result back that essentially says "You have cancer somewhere in your body, we just don't know where"). Lots of smart people, including yours truly, are trying to figure out the best way to evaluate the benefits and harms of MCED tests and to create regulatory frameworks for the companies that are developing them.