Finding Sources of Treatment Variation 

By Brent A. Metfessel MD, MS, CMP™ (Hon)

If a medical provider or healthcare facility is found to have a significant variance from the norm on a measure, such as economic cost, drill-down analysis is important to find the reason behind the variance. 

Episodes of care case-mix adjustment is naturally suited to this kind of analysis, but other population-based groupers such as DCGs also allow drill-down if the clinical categories that are precursors to the assignment of a risk score are used. 

The Concept

The conceptual idea behind drill-down is to obtain greater and greater detail on an area of interest.  Thus, if a provider is found to have a high overall cost variance or performance ratio, a user can select the provider and drill-down into emergency room usage, hospitalization frequency, types of illnesses seen, or procedures performed. 

Case-mix is useful even for the more detailed reports since if, for example, ER use or the utilization of specified procedures is not adjusted for illness burden the “my patients are sicker” argument can easily hold. 

However, if the procedures are related to illness classes, providers can be compared to their peers on procedure use for that illness class.

Example:

Dr. Jones is a family practitioner who had a high patient load from a single large health plan.  These patients under his care had a total of 450 episodes over a two-year period.  His case-mix adjusted performance ratio was 2.28 and cost variance was $157,400.  Dr. Jones requested a drill-down analysis to determine why his practice patterns showed such a high variance from the norm.

One area that the health plan data analysts found had high variance were patients he saw with tendonitis of the lower extremity.  He saw 30 episodes of care for this condition, having a total performance ratio for the illness class of 6.0 and a cost variance of $25,300. 

On further drill-down, the analyst found that the major cost center included the frequency of MRI scans of the lower extremity for the tendonitis patients.  His scan rate was 0.4, which means an average of 4 out of 10 episodes received scans, making a total of 12 scans in all.  His peers of the same specialty showed 0.1 scans per episode of tendonitis of the lower extremity.

Dr. Jones showed a performance ratio of 3.0 and a cost variance of $10,800.  On learning this information, Dr. Jones decided to alter his referral patterns so that his scan rate was brought closer to the norm.

Conclusion

What has been your experience, if any, with drill down analysis; helpful quality improvement adjunct or physician bane?

Related Information Sources:

Practice Management: http://www.springerpub.com/prod.aspx?prod_id=23759

Financial Planning: http://www.jbpub.com/catalog/0763745790

Risk Management: http://www.jbpub.com/catalog/9780763733421

Healthcare Organizations: www.HealthcareFinancials.com

Administrative Terms: www.HealthDictionarySeries.com

Subscribe Now: Did you like this Executive-Post, or find it helpful, interesting and informative? Want to get the latest E-Ps delivered to your email box each morning? Just subscribe using the link below. You can unsubscribe at any time. Security is assured.

Link: http://feeds.feedburner.com/HealthcareFinancialsthePostForcxos

Filed under: Managed Care, Quality Initiatives | 1 Comment »

Measures and Benchmarks

By Brent A. Metfessel MD, MS, CMP™ (Hon)

In a previous Executive-Post, we asked readers if they knew of any case-mix severity measures other than those utilizing expected values.

The Black-Box

When an MCO or HMO analyzes provider practice patterns, it is imperative that the organization educate providers on the methodology and validation of the adjuster, since provider buy-in to the adjuster cannot be obtained otherwise. 

Such education may consist of readings provided with the distributed performance reports that explain the algorithm as well as evidence for the algorithm’s validity.  The MCO needs to be open to questions from providers and show willingness to open the “black box” as much as possible.

Further Considerations

There are further considerations that are relevant to providers when dealing with case-mix adjusted reports:

1. Are the reported performance measures adjusted by specialty? 

The rationale for the additional adjustment comes from the fact that even though a number of specialties may treat congestive heart failure, for example, an internist or family practitioner generally treats less severe cases than would a cardiologist. 

Thus, even if a report is case-mix adjusted by illness class, the adjuster may not fully account for the differences in patient acuity within the illness class.  Adjusting by specialty will enable a more “apples to apples” comparison and achieve greater provider buy-in to the process.

However, for less common illnesses the additional specialty adjustment may cause the cell sizes to become too small, causing the adjustment to lose meaning since there would not be enough patients in some cells for meaningful comparisons.

Overall, whether or not specialty should be added as an additional adjustment is an individual decision made by the health plan.  The larger the health plan, the less chance that cell group sizes may become too small and the greater the advantage of the additional specialty adjustment.

2. What are the exclusion criteria? 

After the case-mix adjustment is performed, it is important that prior to reporting there exists an outlier exclusion criteria.  Without such criteria, there is a much greater chance that a good provider may perform poorly on a performance report since a few high-cost outliers, which may occur due to no fault of the provider, can strongly skew the case-mix indices and lead to artificially high cost variances and performance ratios. 

Some methodologies exclude general catastrophic cases, such as members with costs above $25,000, or there may be a truncation calculation where catastrophic members are included in the reporting information but are truncated to the criteria amount.

Thus, if a patient has costs of $50,000, the costs will be truncated to $25,000 prior to reporting.  This has the advantage of including all patients but the disadvantage of not knowing the actual cost of the patient panel. 

What about Outliers?

Another way to exclude medical outliers involves excluding them at the case-mix class level.  This means that illnesses that generally use less resources will have different criteria – in this case a lower high outlier exclusion boundary – that would an illness class that typically has high resource use. 

If cost is used as the measure of interest, the distribution curve of cost for a particular illness is skewed to the high side and thus does not look like the bell-shaped normal distribution.  This makes developing proper exclusion criteria more complex. 

For greater accuracy, a “non-parametric” or “distribution-free” test is useful.  One such test was developed in 1993 by Sprent and consists of the following equation:

                                 (| X_i – M | / MAD) > Max                                   

Where X_i represents any value being evaluated for outlier status, M represents the median (the value for which 50% of sample values are above, 50% below) of the sample (such as all cases in a disease class) and MAD is the median absolute deviation.

To calculate the MAD value, first obtain the absolute value of the difference between each value and the sample median. Then, sort the difference scores in ascending order. The median of the difference scores is the MAD value. Max is then the criteria point for excluding outliers.

A reasonable value of MAD would be 5.  Both low and high outliers would be excluded based on this equation.

Assessment

Ironically, medical outliers may contain very useful information in themselves. Yet, even more ironically, they are often rejected.

Conclusion

Do you still report outliers separately since such patients, particularly high outliers, may in some cases be steered to case management protocols?

Related Information Sources:

Practice Management: http://www.springerpub.com/prod.aspx?prod_id=23759

Financial Planning: http://www.jbpub.com/catalog/0763745790

Risk Management: http://www.jbpub.com/catalog/9780763733421

Healthcare Organizations: www.HealthcareFinancials.com

Administrative Terms: www.HealthDictionarySeries.com

Subscribe Now: Did you like this Executive-Post, or find it helpful, interesting and informative? Want to get the latest E-Ps delivered to your email box each morning? Just subscribe using the link below. You can unsubscribe at any time. Security is assured.

Link: http://feeds.feedburner.com/HealthcareFinancialsthePostForcxos

Filed under: Managed Care, Quality Initiatives | Leave a comment »

The Centerpiece of Quality Practice Patterns

By Brent A. Metfessel MD, MS, CMP™ (Hon)

It is difficult to construct an adequate medical practice pattern profile without case-mix or risk adjustments. There needs to be an algorithm that adjusts for the medical severity of patient mix. 

For example, a tertiary care center in New York City cannot be compared using unadjusted data with a community hospital outside the city. The tertiary care center will use more resources, and thus cost more, than the community hospital no matter how exemplary the tertiary care center. 

And, a cardiologist cannot be compared to a family practitioner, since in general the cardiologist will see patients of greater severity. 

Algorithms for Case-Mix Adjustment

A wide variety of methodologies exist that are useful for case-mix, risk, and severity of illness adjustment. And, a number of third-party vendors exist that sell software groupers for case-mix categorization.

Since each methodology has different strengths, some MCOs have purchased more than one software package. There is no such thing as a “perfect” adjuster. Five examples of commonly used algorithms follow:

·         Diagnosis Related Groups (DRGs) and related adjusters: Originally put into use in the early 1980s, DRGs were intended for use mainly as a methodology for Medicare to determine reimbursement for hospital stays.  Nevertheless, DRGs and their more recent derivatives (Revised DRGs or RDRGs, and All Patient Refined DRGs or APR-DRGs, both of which subclassify each DRG category into three to five severity strata using various algorithms) are useful for inpatient case-mix adjustment.  An example of a DRG category is DRG 89, “Simple pneumonia & pleurisy, age > 17, with CC [complications]”. The same can be said adjusters related to the newest Medical Severity DRGs [MS-DRGs].

·         Episode Treatment Groups™ or ETGs (Symmetry Health Data Systems, Inc.): This data grouper classifies the claims records into episodes of care that track the progress of an acute illness from onset to resolution and includes related diagnoses and treatments.  For more chronic illness episodes, where there is really no defined “onset” or “resolution”, one usually profiles providers on a pre-defined time window, such as a year-long episode. To capture enough episodes for analysis, ETGs generally require a two-year reporting period. Since this case-mix adjuster depicts the longitudinal aspects of care, ETGs are a process-based adjuster, meaning that they emphasize the process of care and the treatment the patient receives over a time course. A member can, and often does, have more than one ETG during a reporting period. An example of an ETG is “Obesity, morbid, with surgery”. There exist over 600 ETG categories, which are granular enough to detect nuances in illness classes and severity but not so large as to lead to significant small cell size problems. ETGs also group pharmacy claims and attach them to the most relevant episode based on priority tables. Over 400 health plans have purchased the grouper as of May, 2003, and 700 by 2008. In addition, Episode Risk Groups™, a derivative of ETGs, can be used prospectively for predictive modeling of cost as well.

·         Adjusted Clinical Groups or ACGs (Johns Hopkins University): ACGs group illnesses into morbidity clusters rather than specific diseases as do ETGs. An example of an ACG is “Acute major and likely to recur”. Since ACGs are based on morbidity clusters, patients with multiple complex illness conditions can be readily identified.  Since each patient has only one ACG for an entire reporting period, such an adjuster is called population-based.  The process of care over time is not as important with such algorithms. In fact, ACGs do not require procedure or CPT codes at all – just ICD diagnoses, age, gender, and member and provider identification fields, which gives the methodology the advantage of input simplicity. There exist over 100 ACGs at present, and they are in use at nearly 200 organizations worldwide. In general there are fewer categories in population-based adjusters than in process-based adjusters, since process-based algorithms need to account for specific diseases.

·         Diagnosis Cost Groups™ or DCGs (DxCG, Inc.):  DCGs are also a population-based grouper.  Although the grouper begins with 184 Condition Categories (ex: “Benign neoplasm of skin”). These Condition Categories are also sorted into hierarchies and aggregated into broader categories. The combinations of Condition Categories that a member has can then be used to predict health care resource utilization based on an overall risk score for each member. This prediction can either be for the current year or for the subsequent year, depending on the model used. Over 100 organizations now use DCGs, and like ACGs they do not require procedure codes. One important feature of DCGs is its ability to be used in predictive modeling of prospective resource use, using a different model than that used for retrospective analysis

·         Age-gender:  In these models, various age and gender strata are used to account for risk.  Generally there are about 9 to 20 strata for age gender, depending on the needs of the health plan. Basically, resource use is moderate in the early years up until about age 5, then decreases through adolescence and the 20s, then slowly rises again in a non-linear fashion until it becomes quite high in the senior years. Females also tend to use more resources during their reproductive years. Of all the models described, age-gender has the least explanatory power for the prediction of resource utilization either retrospectively or prospectively. The ability of a case-mix adjuster to explain variation in resource utilization is determined by the “R-squared” (the square of the correlation coefficient), with the case-mix categories or risk score as the independent variables and a measure of resource use (such as cost) as the dependent variable. Age-gender models have an explanatory power of about 3 – 7% while publications on proprietary adjusters have generally shown that they explain about 30 – 50% of the variation for retrospective analysis. Prospective explanatory power is somewhat less, usually around 15 – 25%.

Assessment

Medical providers have the right to ask that reports dealing with health care resource utilization have proper case-mix or severity of illness adjustment, and that resources are available at the health plan or MCO to answer questions concerning the adjustment algorithm and to offer a complete explanation of the case-mix methodology used.

Conclusion

Many MCOs and HMOs now provide literature to physicians and medical providers that discuss the reporting and case-mix methods when the profile reports are distributed. Are you aware of them; please comment and opine on their use, or abuse? 

Related Information Sources:

Practice Management: http://www.springerpub.com/prod.aspx?prod_id=23759

Financial Planning: http://www.jbpub.com/catalog/0763745790

Risk Management: http://www.jbpub.com/catalog/9780763733421

Healthcare Organizations: www.HealthcareFinancials.com

Administrative Terms: www.HealthDictionarySeries.com

Subscribe Now: Did you like this Executive-Post, or find it helpful, interesting and informative? Want to get the latest E-Ps delivered to your email box each morning? Just subscribe using the link below. You can unsubscribe at any time. Security is assured.

Link: http://feeds.feedburner.com/HealthcareFinancialsthePostForcxos

Filed under: Managed Care, Quality Initiatives | Leave a comment »