How We Compiled The Dollars For Docs Hospital Data
Our goal was to compare U.S. hospitals based on the percentage of their affiliated physicians who receive payments of various sizes from pharmaceutical and medical device companies.
To do this, we used several databases.
The data we used
First, for information on industry payments to doctors, we turned to the federal government’s Open Payments system. All companies are required to submit this data, which is then made public under the Physician Payment Sunshine Act, a part of the 2010 Affordable Care Act. We used a file called “General payments,” which includes categories like promotional speaking, consulting, meals, travel, gifts and royalties. It does not include payments for research. The data we used covers calendar year 2014 and includes records on 605,680 providers (doctors, dentists, chiropractors, podiatrists and optometrists).
Second, to determine which physicians practiced at which hospitals, we used Medicare’s Physician Compare data, which includes up to five hospital affiliations for each physician. Because 2014 was the only full year for which Open Payments data existed at the time of our analysis, we used archived Physician Compare data, thus providing a snapshot of physicians’ hospital affiliations as of December 2014. This file includes records on 890,707 health providers, of which 600,348 had at least one hospital affiliation listed. Note: This data does not include doctors who don’t participate in Medicare and those who don’t admit many patients to the hospital. Also, an affiliation does not mean that a doctor is employed by the hospital.
For information about a hospital’s characteristics, including address and ownership, we relied on Medicare’s Provider of Services file from 2014, as well as data from the American Hospital Association Annual Survey. Our analysis included 4,815 hospitals. Of those, 2,241 are included inProPublica's toolbecause they had at least 50 affiliated physicians.
Finally, we accessed ratings assigned to selected teaching hospitals in 2014 by the American Medical Student Association, which compiles a tool called the AMSA Scorecard. The group reviewed the conflict-of-interest policies of 204 teaching hospitals and gave the hospitals either an A (the best grade), a B, a C or an Incomplete because their policies were “insufficient for evaluation.” The 2014 data was no longer available on AMSA’s website, so we accessed it via the Internet Archive.
Cleaning and combining the data
Combining these datasets involved many choices, as well as voluminous data cleaning.
Health care providers are assigned to hospitals either by self-reporting or on the basis of their Medicare claims. The first hospital listed denoted the strongest affiliation. ProPublica shared the listed hospital assignments in Physician Compare with several hospitals in an attempt to verify their accuracy.
After consulting with these hospitals, we made a decision that the best approach was to include only each doctor’s primary hospital affiliation (variable name “Claims based hospital affiliation CCN 1” in the Physician Compare file). The hospitals we consulted told us that the lower affiliations contained more errors. We also made a decision to include only the top 100 medical specialties in both the Physician Compare and Open Payments data. This would exclude nurse practitioners and physician assistants, who do not have to report industry payments, and doctors who aren’t enrolled in Medicare. We then excluded the specialty “Hospitals, General Acute Care Hospital,” bringing the total to 99 common specialties. (You can see that list below.)
The Physician Compare dataset uses National Provider Identifier (NPI) numbers to identify physicians, whereas the Open Payments system uses a different unique ID for doctors. In early 2016, ProPublica matched the physicians in the Open Payments data to their NPI numbers. Our methods were able to match more than 99.7 percent of physicians in the Open Payments data to their NPI numbers. You can read more about that process here.
ProPublica then matched the 2014 Physician Compare data with the 2014 Open Payments data, using physicians’ NPI numbers as the matching key. All told, we found 498,453 doctors in the top 99 medical specialties who were affiliated with a primary hospital in the Medicare data. Of those, 329,111 received at least one general payment in 2014 (we excluded research payments.)
For every hospital, we calculated the number of doctors who listed it as the primary affiliation in December 2014, the number of those who received a payment (and percentage), the total number of payments to doctors at that hospital, the number who received at least $100 in payments, the number who received at least $1,000 in payments, the number who received at least $5,000 in payments, and the number who were promotional speakers. For our national totals, we included all hospitals. But for ProPublica's online tool to avoid inflated percentages when comparing hospital payment rates, we only included hospitals with 50 or more physicians listed as primary affiliates.
ProPublica then looked at how physician payment rates varied by hospital ownership. To conduct this analysis, we used data from the American Hospital Association Annual Survey, considered the gold standard by researchers.
To ensure that we made a proper decision to exclude lesser hospital affiliations, we calculated the correlation between payment rates for hospitals using only physicians who had a primary affiliation with the payment rates for hospitals using all listed physician affiliations. We noted a strong correlation at .93.
Calculating statistical significance
To test the statistical significance of the differences in proportion of doctors taking payments by hospital ownership types, we conducted a chi-square test. To identify which pairs of ownership types had statistically significant differences, we conducted post-hoc pairwise chi-square tests with Bonferroni corrections. All pairs were statistically significant.
Using the same methods, we also tested the significance of differences in proportion of doctors taking payments by the American Medical Student Association hospital rating. All pairs were statistically significant. It’s worth noting that AMSA rated 204 teaching hospitals. Our analysis included only 152 of those hospitals; almost all of the rest were hospitals run by the U.S. Department of Veterans Affairs, which were not included in our data.
List of specialties analyzed
Obstetrics & Gynecology
Hematology & Oncology
Anatomic Pathology & Clinical Pathology
Student in an Organized Health Care Education/Training Program
Endocrinology, Diabetes & Metabolism
Physical Medicine & Rehabilitation
Podiatrist, Foot & Ankle Surgery
Thoracic Surgery (Cardiothoracic Vascular Surgery)
Critical Care Medicine
Clinical Cardiac Electrophysiology
Vascular & Interventional Radiology
Orthopaedic Surgery of the Spine
Allergy & Immunology
Colon & Rectal Surgery
Plastic and Reconstructive Surgery
Interventional Pain Medicine
Maternal & Fetal Medicine
Emergency Medical Services
Dentist, Oral and Maxillofacial Surgery
Podiatrist, Foot Surgery
Adult Reconstructive Orthopaedic Surgery
Clinical Pathology/Laboratory Medicine
Child & Adolescent Psychiatry
Surgical Critical Care
Otolaryngology/Facial Plastic Surgery
Critical Care Medicine
Foot and Ankle Surgery
Hospice and Palliative Medicine
Neuromusculoskeletal Medicine & OMM
Podiatrist, Primary Podiatric Medicine
Plastic Surgery within the Head & Neck
Neurology with Special Qualifications in Child Neurology
Female Pelvic Medicine and Reconstructive Surgery
Facial Plastic Surgery
Oral & Maxillofacial Surgery