The trusted source for
healthcare information and
By Patrice Spath, RHIT
Forest Grove, OR
Many hospitals are comparing the practice patterns of physicians. The simplest way to profile physicians’ hospital practices is to rank them by their patients’ length of stay (LOS) and hospital charges. The assumption is that physicians whose patients have a higher LOS and/or charges are the physicians who have high complication rates (e.g., wound infections, etc.), provide less than adequate patient care (e.g., inappropriate antibiotic therapy leading to slower recovery), or are not practicing cost-effective care (e.g., inadequate planning or performance).
The major problem encountered when judging physicians’ average LOS or patient charges is that these data fail to take into account patient severity of illness. Some physicians may very well care for sicker patients as compared to other physicians and thus have appropriately higher LOS and charges.
The underlying assumption of any method used to adjust for patient severity of illness is that resource use results from a complex mix of factors. In order to control for patient severity of illness, risk adjustors must be applied to the data. The difficulty is choosing the risk factors to use.
A risk adjustor that predicts one outcome (e.g., extended hospital stay) may not predict another outcome (e.g., use of high-cost resources). Many diverse patient attributes affect risks, including age, sex, acute physiological stability, reason for hospitalization and severity of the condition, the extent of comorbid illnesses, functional status, psychosocial and cultural factors, socioeconomic characteristics, and patient preferences.
There are a large number of methodologies for adjusting physician profiles according to the severity of their patients’ illness. Many hospitals use discharge abstract-based severity measures such as All Patient Refined DRGs (APR-DRGs) or Iameter’s Acuity Index Method (AIM) for risk-adjustment. These systems rely only on data found in the discharge database to risk adjust patients’ severity of illness. The amount of information in this database is limited (e.g., diagnoses and procedures coded using ICD-9-CM; admission source; and discharge disposition). However, despite limited clinical information, discharge data offer the advantages of uniformity, availability, and computer readability.
Some methods for risk adjusting patient populations use clinical data abstracted from patient records. Two commonly used systems are MedisGroups and the Acute Physiology and Chronic Health Evaluation (APACHE). These systems may be more clinically credible, but the cost of gathering additional information from patient records can be an impediment. Whether the additional cost is worth the effort continues to be hotly debated.
Case managers should know whether the physician profile data used to examine utilization patterns in their hospital are risk-adjusted. If physician profile data are not risk-adjusted, it may be impossible to draw strong inferences about inappropriate utilization practices based solely on LOS and charge data.
If the data are risk-adjusted, case managers should be aware of the factors that are used to arrive at the patient severity scores and what limitations may exist. For example, most severity measures do not include all patient characteristics that increase risk, such as physical functional status, patients’ preferences for care and outcomes, cultural factors, and socioeconomic characteristics.
Not all assessments of health care quality and cost-effectiveness need to include a severity of illness measure. For instance, data showing physicians’ rate of compliance with clinical practice guideline recommendations do not need to be risk adjusted. Listed below are examples of measures that evaluate practice-related issues:
— percent of patients with myocardial infarction for whom aspirin is ordered or the rationale for nonuse is documented by the physician;
— percent of patients with congestive heart failure for whom an ACE (angiotensin-converting enzyme) inhibitor is ordered or the rationale for nonuse is documented;
— percent of patients undergoing total hip replacement who receive prophylactic antibiotics within the two-hour window prior to surgery or the rationale for nonuse is documented;
— percent of patients with cerebrovascular accident who are initially evaluated with a CT scan (not the more expensive MRI scan).
There is convincing evidence that both quality and cost-effectiveness suffer when recommendations found in evidence-based guidelines are not followed — and the validity of these measures does not depend upon differences in patients’ severity of illness.
Monitoring of possibly avoidable hospital days also does not require a severity of illness measurement system. A possibly avoidable day is independent of the patient’s severity of illness and is determined instead on the patient’s medical stability.
Even very ill patients can be adequately managed at nonhospital, lower levels of care if they are medically stable. For example, each one of the following cases could be cared for safely at a lower level of care such as a skilled or subacute facility or at home with home care agency visits:
The data gathered by case managers about possibly avoidable hospital days could show important differences in physicians’ practice patterns. Adjusting the data for patient risk factors is not necessary. Some physicians are reluctant to accept the validity of quality and cost-effectiveness data if they are not risk-adjusted. When profiles show their patients to have higher costs or longer lengths of hospital stay, many will say, "It’s because I have sicker patients than my colleagues." For some quality and cost measures, risk-adjusted data can be an important adjunct. However, there are many performance measures that can be used on physician profiles that are not affected by patients’ severity of illness.