Physician Spending And Its Association With Patient Outcomes
Global spending on health is soaring at a rapid rate. The total health expenditure (THE) worldwide increased from US$ 7.6 trillion in 2016 to US$ 7.8 trillion in 2017. The World Health Organization 2020 report stated that it reached US$ 8.3 trillion or 10% of global GDP. Among western countries, the United States spent approximately 17.7% of GDP on health in 2019 and is projected to reach 19.7% GDP in 2028. This is in comparison with 9-10% in Western European countries and 7-8% in OECD states. Nonetheless, the Commonwealth Fund report stated that Americans experience worse health outcomes than their peers in most countries. For example, lowest life expectancy, highest suicide rates, and highest rates of avoidable mortality.
While health expenditure depends on factors outside the health system such as social, economic, and political factors, evidence suggests that about one-quarter of total health care spending in the United States is wasteful. Furthermore, a few studies have documented significant variation in spending and patient outcomes across physicians.
Is there a real value that can be obtained from this higher physician spending? A few authors tried to examine higher physician spending and its association with patient outcomes. One such study was a retrospective data analysis published in JAMA Internal Medicine.
The investigators used a 20% random sample of Medicare patients (65 years and older) hospitalized with a general medical condition between January 1, 2011, and December 31, 2014, and treated by a hospitalist or general internist. Patients treated in acute care hospitals were included whereas elective hospitalizations and hospitalizations in which patients left against doctor’s advice were excluded. To ensure sufficient follow-up, patients admitted in December 2014 for 30-day mortality analyses and patients discharged in December 2014 for readmission analyses were also excluded.
To avoid unstable estimates of physician spending, the sample was restricted to physicians with at least 10 observed hospitalizations between 2011 and 2012. Ordinary least squares models were used to measure Part B physician spending per hospitalization. Hospitalizations in the top and bottom 5 percent of residuals were excluded to address outliers. Subsequent diagnostics established that the statistical model fits the data with respect to linearity, homoscedasticity, and normality. Next, the observed-to-expected spend ratio for each physician was calculated. Standardized spending level for each physician was obtained by multiplying these ratios by the grand mean of spending per hospitalization
Variables that could influence spending and the two patient outcomes, mortality and readmissions were adjusted. Patient characteristics included age, sex, race or ethnicity, MS-DRG, 27 coexisting chronic conditions, median household income, and whether patients had Medicaid coverage. Physician characteristics included: age, sex, the medical school graduated from, and type of medical training (allopathic vs osteopathic).
To analyze the total variance in spending in hospital-, physician-, and hospitalization-levels, a cross-classified multilevel model was developed. This statistical model allowed the investigators to divide total variation into different levels, even for the physicians practicing in multiple hospitals.
Variance Partition Coefficients (VPCs) and Inter Unit Reliability (IUR) were used to describe the percentage of variation in standardized part B spending explained by hospitals, physicians, and patients.
Sensitivity analyses conducted were as follows:
- Analyses using the sum of Part A and B spending
- Alternative attribution methods – (i) by attributing patients to physicians with the largest number of evaluation and management claims, and (ii) attributing patients to admitting physicians who billed the first evaluation and management claim for a particular hospitalization.
- Addressing potential outlier cases by (i) trimming the top and bottom 3% of residuals, and (ii) trimming the top and bottom 7% of residuals.
- Sensitivity threshold of 30 to assess if the findings were sensitive to the threshold of 10 hospitalizations per physician for the samples.
- At times physicians who consistently transfer their sickest patients might be misclassified as low-spending physicians. To address this, all patients transferred to acute care hospitals and allocated subsequent spending to the initial hospital were identified.
- Exclusion of cancer patients and those discharged to hospice as they are strong predictors of do-not-resuscitate (DNR) directives.
- Length of stay (LOS) adjustments as it may capture residual confounding.
- Sensitivity analysis of overall general internists instead of hospitalists alone.
- Evaluation of medical conditions such as sepsis, pneumonia, congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), and urinary tract infection (UTI).
SAS, version 9.4 (SAS Institute) and Stata, version 14 (StataCorp) was used for all analyses.
RESULTS & DISCUSSION
The final sample for the analysis of hospitalists consisted of 485,016 hospitalizations, treated by 21,963 hospitalists, in 2837 hospitals, across the country. In the case of general internists, the final sample consisted of 839,512 hospitalizations, treated by 50,079 physicians, in 3195 hospitals.
For hospitalists, 8.4% of the total variation in health care spending was attributed to differences between individual physicians, compared with 7% explained by differences between hospitals. For general internists, it was 10.5% across physicians vs 6.2% across hospitals. Meaning, the variation in health care spending was more across individual physicians within the same hospital than across hospitals. The IUR of 0.73 for hospitalists (considering the average number of hospitalizations being 27.4) indicates that the variation in spending between physicians was due to differences between physicians.
It was also demonstrated that within the same hospital, for similar patients the highest vs lowest spent $1055 vs $743 i.e per hospitalization, the highest spending physicians spent about 40% more than the lowest-spending physicians.
The overall 30-day mortality rate and readmission rate were 11.0% and 14.5%, respectively. The authors observed no association between higher spending and mortality or readmission rates; the same results were found when the analysis was extended to general internists.
The above results indicate that health care policy reforms should also take doctors into account, rather than solely focusing on hospitals.
There are several limitations to this study.
- First, it calculated only the physician-level estimates of Part B spending; the estimates from a larger patient sample seen by each physician could differ.
- Second, in Medicare claims, inpatient medications, laboratory tests, and imaging data are included in the fixed Part A payment. This precluded the authors from measuring how these items vary across physicians.
- Third, this study measured only two of the several healthcare outcome measures and excluded top measures such as patient experience and safety of care.
- Last, this retrospective analysis was restricted to the hospitalized Medicare patients with medical conditions, and therefore, cannot be generalizable to non-Medicare patients, patients with surgical conditions, or the ones in ambulatory care.
Globally, health care costs have been rising for decades and are expected to keep increasing. In the U.S, the spending varies more across individual physicians, even within the same hospital. However, that greater spending does not lead to improvement in patient outcomes.
Though it’s too early to say whether policies targeting physicians may be more effective in reducing wasteful health care spending, this study results highlight the need for meaningful, efficient, as well as effective health care reforms in the U.S.
Oncology Related Tools
- Prognostic Scoring for Myelofibrosis
- Opioid Conversion Calculator
- Updated Advanced Opioid Conversion Calculator
- Nonsteroidal anti-inflammatory drugs (NSAID) Selection Tool
- Absolute Neutrophil Count Calculator
- Body Surface Area (BSA) Multi-Calc
- Carboplatin AUC Calculator
- Carboplatin AUC – Updated Version
- Urinary Indices, Renal Failure Index (RFI) and Fractional Excretion of Sodium (FE-NA)
- Creatinine Clearance (CRCL) – Standard Calculator
- Creatinine Clearance Multi-Calc – All of the latest research
- Patient Controlled Analgesia (PCA) Settings
- Intravenous Antineoplastic Agents – Administration Guidelines
- Therapeutic Drug Levels
- Beers Criteria for potentially inappropriate medications
- Allergic response? 12-step desensitization protocol
- Protein requirements calculator
- Basal Metabolic Rate (BMR) Multi-calc (Estimate caloric requirements)
- Irritable Bowel Syndrome Treatment Options
- Common Anti-emetics
- Fall Assessment – Berg Balance Scale
- Physician Burnout: Causes and Prevention
- LEAP-MS: Adaptations for Advanced Stages
- MS: Exercise Impacts on MRI
- The Role of Preretirement Job Complexity in Cognitive Performance
- Extrapontine Myelinolysis and PTA in Pregnancy
- Verbal Communication and Masks
- Sugammadex Versus Neostigmine in Thyroidectomy
- SGLT Inhibitors on Weight and Lipid Metabolism in Diabetes
- Saxagliptin: Obese Patients with Impaired Glucose Tolerance
- Levothyroxine Therapy and Depression
- Grave’s Disease and Risk of Systemic Lupus Erythematosus
- Benign Thyroid Removal and Patient Satisfaction
- MF- Biology, Management, and a Case Study of Ocular Manifestation
- Quality Of Life In Adolescent Cancer Survivors
- Cancer Opioid Risk Score
- Oncology-Specific Opioid Risk Calculator In Cancer Survivors
- 3D MRI for Non-invasive Ocular Proton Therapy of Uveal Melanomas
- Sexual Dysfunction in Prostate Cancer Patients
- 3-Day Surprise Question To Predict Survival Rates in Advanced Cancer Patients