Patient Sociodemographics Affect Physician Quality Scores

TORONTO — Physician practices treating higher proportions of less-educated patients have consistently lower performance scores, reported Dr. Mark Friedberg of the division of general medicine at the Brigham and Women's Hospital and the Harvard School of Public Health, both in Boston.

In fact, an increase of one standard deviation in the proportion of non-college graduate patients is associated with a significant Health Employer Data and Information Set (HEDIS) performance score decrease of as much as 2.5%.

“Our concern is that practice sites caring for disproportionate shares of vulnerable patients may be penalized by public performance reporting and pay-for-performance contracts,” said Dr. Friedberg who presented preliminary research at the annual meeting of the Society of General Internal Medicine.

“Without adjusting HEDIS scores for patient sociodemographic characteristics—or adjusting some aspect of the way these scores are used—physicians may feel an incentive to avoid patients from vulnerable populations,” he said.

The measurement of primary care quality for public reporting has become a hot issue in recent years, with physicians who care for minority patients and those with lower incomes worried that they may be at a disadvantage in a system with a one-size-fits-all approach to quality measurement.

Dr. Friedberg noted a recent study (Health Aff. 2007;26:w405-w414 [Epub doi:10.1377/hlthaff.26.3.w405]) that found that 85% of physicians polled agreed with the statement: “At present, measures of quality are not adequately adjusted for patients' socioeconomic status.”

Fully 82% were concerned that measuring quality may deter physicians from treating high-risk patients.

Dr. Friedberg and his colleagues used the Massachusetts Health Quality Partners (MHQP) statewide reporting program, which supplied data from commercial insurers aggregated at the physician level on eight HEDIS measures: breast cancer, cervical cancer, chlamydia, asthma controller medications, HbA_1c testing, cholesterol testing, eye exams, and nephropathy.

MHQP is a statewide collaborative that includes the five largest health plans in Massachusetts, contracting with 90% of state primary care providers and covering 63% of Massachusetts residents, or about 4 million people.

Data were collected from 241 physician practice sites (including 1,489 physicians) that provided adult primary care to insured patients during 2004.

These data were linked to patient responses from the 2002–2003 Massachusetts Ambulatory Care Experiences Survey to calculate the prevalence of sociodemographic characteristics (age, gender, race, ethnicity, and education) within each practice site's patient panel. Practice site was used as the unit of analysis.

Median site-level HEDIS scores ranged from 94% for HbA_1c screening to 43% for chlamydia screening in women aged 21–25 years.

In bivariate analyses, lower site-level proportions of college graduate patients were significantly associated with lower HEDIS scores on all eight measures. These associations remained statistically significant for seven of the eight measures even after multivariate adjustment. Significant bivariate associations between sites' HEDIS scores and the age, racial, and ethnic composition of their patient panels existed for chlamydia screening, but these ties did not remain statistically significant after multivariate adjustment.

TORONTO — Physician practices treating higher proportions of less-educated patients have consistently lower performance scores, reported Dr. Mark Friedberg of the division of general medicine at the Brigham and Women's Hospital and the Harvard School of Public Health, both in Boston.

In fact, an increase of one standard deviation in the proportion of non-college graduate patients is associated with a significant Health Employer Data and Information Set (HEDIS) performance score decrease of as much as 2.5%.

“Our concern is that practice sites caring for disproportionate shares of vulnerable patients may be penalized by public performance reporting and pay-for-performance contracts,” said Dr. Friedberg who presented preliminary research at the annual meeting of the Society of General Internal Medicine.

“Without adjusting HEDIS scores for patient sociodemographic characteristics—or adjusting some aspect of the way these scores are used—physicians may feel an incentive to avoid patients from vulnerable populations,” he said.

The measurement of primary care quality for public reporting has become a hot issue in recent years, with physicians who care for minority patients and those with lower incomes worried that they may be at a disadvantage in a system with a one-size-fits-all approach to quality measurement.

Dr. Friedberg noted a recent study (Health Aff. 2007;26:w405-w414 [Epub doi:10.1377/hlthaff.26.3.w405]) that found that 85% of physicians polled agreed with the statement: “At present, measures of quality are not adequately adjusted for patients' socioeconomic status.”

Fully 82% were concerned that measuring quality may deter physicians from treating high-risk patients.

Dr. Friedberg and his colleagues used the Massachusetts Health Quality Partners (MHQP) statewide reporting program, which supplied data from commercial insurers aggregated at the physician level on eight HEDIS measures: breast cancer, cervical cancer, chlamydia, asthma controller medications, HbA_1c testing, cholesterol testing, eye exams, and nephropathy.

MHQP is a statewide collaborative that includes the five largest health plans in Massachusetts, contracting with 90% of state primary care providers and covering 63% of Massachusetts residents, or about 4 million people.

Data were collected from 241 physician practice sites (including 1,489 physicians) that provided adult primary care to insured patients during 2004.

These data were linked to patient responses from the 2002–2003 Massachusetts Ambulatory Care Experiences Survey to calculate the prevalence of sociodemographic characteristics (age, gender, race, ethnicity, and education) within each practice site's patient panel. Practice site was used as the unit of analysis.

Median site-level HEDIS scores ranged from 94% for HbA_1c screening to 43% for chlamydia screening in women aged 21–25 years.

In bivariate analyses, lower site-level proportions of college graduate patients were significantly associated with lower HEDIS scores on all eight measures. These associations remained statistically significant for seven of the eight measures even after multivariate adjustment. Significant bivariate associations between sites' HEDIS scores and the age, racial, and ethnic composition of their patient panels existed for chlamydia screening, but these ties did not remain statistically significant after multivariate adjustment.

TORONTO — Physician practices treating higher proportions of less-educated patients have consistently lower performance scores, reported Dr. Mark Friedberg of the division of general medicine at the Brigham and Women's Hospital and the Harvard School of Public Health, both in Boston.

In fact, an increase of one standard deviation in the proportion of non-college graduate patients is associated with a significant Health Employer Data and Information Set (HEDIS) performance score decrease of as much as 2.5%.

“Our concern is that practice sites caring for disproportionate shares of vulnerable patients may be penalized by public performance reporting and pay-for-performance contracts,” said Dr. Friedberg who presented preliminary research at the annual meeting of the Society of General Internal Medicine.

“Without adjusting HEDIS scores for patient sociodemographic characteristics—or adjusting some aspect of the way these scores are used—physicians may feel an incentive to avoid patients from vulnerable populations,” he said.

The measurement of primary care quality for public reporting has become a hot issue in recent years, with physicians who care for minority patients and those with lower incomes worried that they may be at a disadvantage in a system with a one-size-fits-all approach to quality measurement.

Dr. Friedberg noted a recent study (Health Aff. 2007;26:w405-w414 [Epub doi:10.1377/hlthaff.26.3.w405]) that found that 85% of physicians polled agreed with the statement: “At present, measures of quality are not adequately adjusted for patients' socioeconomic status.”

Fully 82% were concerned that measuring quality may deter physicians from treating high-risk patients.

Dr. Friedberg and his colleagues used the Massachusetts Health Quality Partners (MHQP) statewide reporting program, which supplied data from commercial insurers aggregated at the physician level on eight HEDIS measures: breast cancer, cervical cancer, chlamydia, asthma controller medications, HbA_1c testing, cholesterol testing, eye exams, and nephropathy.

MHQP is a statewide collaborative that includes the five largest health plans in Massachusetts, contracting with 90% of state primary care providers and covering 63% of Massachusetts residents, or about 4 million people.

Data were collected from 241 physician practice sites (including 1,489 physicians) that provided adult primary care to insured patients during 2004.

These data were linked to patient responses from the 2002–2003 Massachusetts Ambulatory Care Experiences Survey to calculate the prevalence of sociodemographic characteristics (age, gender, race, ethnicity, and education) within each practice site's patient panel. Practice site was used as the unit of analysis.

Median site-level HEDIS scores ranged from 94% for HbA_1c screening to 43% for chlamydia screening in women aged 21–25 years.

In bivariate analyses, lower site-level proportions of college graduate patients were significantly associated with lower HEDIS scores on all eight measures. These associations remained statistically significant for seven of the eight measures even after multivariate adjustment. Significant bivariate associations between sites' HEDIS scores and the age, racial, and ethnic composition of their patient panels existed for chlamydia screening, but these ties did not remain statistically significant after multivariate adjustment.