Outcomes Terms & Ideas

1.) Cohort- A designated group of patients. For example, all patients over 50 undegoing a TKA is a cohort.

2.) Cohort Study- A study that follows a cohort longitudinally in time. The study may be prospective or retrospective.

3.) Prospective Cohort Study- The prospective study begins right when the patient first sees the treating physician. A rigid protocol determining what information is to be gathered, how it is to be gathered and how it is going to be analyzed is already in place. This includes the criteria on which patients may enter the study or be excluded from the study, what surgical procedure is to be performed, how long the follow up will be and when follow ups will take place. Thus, the reliability and validity of a prospective study are superior to a retrospective study. All modes of collecting data are designed specifically for the study being done where as in a retrospective study there is the potential to find missing information and faulty data due to the fact that the data was not collected with the specific study in mind. Prospective studies allow comparisons between pre-and-post operative status. In many retrospective studies, patients are asked how they felt before they had their surgery. This leads to obvious problems with the pre-operative data collected in retrospective studies. The fact that the patient has already undergone the procedure in question immediately introduces bias into the equation and thus the validity of the study must be questioned. At the present time, it is essential that all Outcomes projects be conducted prospectively if we are to maintain scientific validity. That is not to say that prospective studies are the gold standard of research. They are full of bias which brings the question of validity to the forefront. They are, however, superior to retrospectively conducted studies.

4.) Retrospective Cohort Study- If the study being performed is started after the results have taken place, then the study is retrospective. Traditionally, most of the medical literature to this date has been done retrospectively. Yet, it has become evident that many problems with this type of study diminish the validity and accuracy of the results. Such problems include: having to rely on medical records that were not necessarily prepared for outcomes studies; relying on the patient's ability to recall information and past sensations to gather needed information not to mention treatment that may modify a patient's perception as to how he/she felt prior to that treatment. Furthermore, the outcome in question has already taken place so the physician cannot really determine what outcomes to use. For these reasons, the retrospective study is becoming obsolete.

5.) Case-Controlled Studies- Studies in which one group of patients with the outcome in question, as well as one group without the outcome in question, are gathered and looked at backward in time to assess the results of a particular treatment. Case controlled studies are retrospective. However, they are more valuable than a retrospective cohort study due to the presence of a control group which increases the scientific validity of the study. Either a historical or concurrent control group may be used with preference given to concurrent control groups for obvious reasons.

6.) Randomized Clinical Trials- The RCT was developed to deal with the bias problems in the prospective cohort studies. The RCT is a prospective concurrently controlled cohort study that includes every component of a propspective cohort study. The difference between a RCT and PCS is that the treatment in question allocated randomly to the cohort groups. The main goal of this randomization is to eliminate susceptibility bias. For example, let's say 20 obese patients are undergoing a particular type of knee surgery. The surgeon wants to discover if one procedure is better than another. Unfortunately, he puts all 20 obese patients in one cohort group. This obviously makes the two cohorts different, thus comparing them is scientifically irrelevant because the obesity of group one will affect the outcome of that procedure. By randomizing patients to one of two treatment groups this susceptibility bias does occur because presumably about half the obese patients will receive treatment 1 and half will receive treatment 2. Three other types of bias that may occur are: performance bias, detection bias, and transfer bias. Performance bias results when the surgeon in question is more experienced with procedure A than with procedure B. Thus, performance bias results from dissimilar levels in skill in performing the procedures. Detection bias occurs when the outcomes in question are not measured in a comparable way. This means that the outcome tools used must be used in both treatments. Transfer bias results from differential loss to follow up in the treatment groups. The ideal way to deal with performance and detection bias is by double blinding. In double blinding, both the assessor and the patient are unaware of which treatment the patient has received. However, double blinding is unrealistic in a surgical study as the surgeon surely knows which patients have had which particular procedure. Secondly, the ethical issues are involved as patients may be concerned that do not know what type of treatment they will be receiving. So a single blind study is the next best alternative where someone other than the surgeon assesses the results of the procedure. The third type of bias, transfer bias, can only be prevented by maintaining as complete a follow up as possible. When loss of follow up does occur, the reasons must be given for the loss. The reason the literature is devoid of RCTs is due to the many ethical dilemmas they create. Physicians want to give their patients the best possible care, but RCTs require that in some instances patients be given an inferior mode of treatment for the benefit of future patients. It is difficult for physicians to knowingly allocate an inferior treatment. Secondly, the "catch 22" of RCTs is the new procedures can't be studied until they have been tested on patients. This creates quite a dilemma. The answer to some of these problems may be solved by the Random Surgeon design. In this instance, surgeons are assigned to either treatment group based on their experience and expertise with the procedure. This allows surgeons to remain committed to their favored procedure thus eliminating some of the problems with assigning patients to a treatment that the surgeon believes is inferior. The main idea of the Random Surgeon design is maintain the surgeon-patient relationship that the RCTs disrupt.

7.) Outcome Measures- The tools by which outcomes are collected and measured can be catergorized in three in areas: Clinical (or Process) Measures, health status measures, and patients satisfaction measures. Clinical Measures are the objective data that the physician obtains by physical exam. This can include, range of motion, radiographic information, and other such measurable data. Examples are the Harris Hip Score and ASES Total Shoulder Arthroplasty Score. Health Status measures are usually in the form of questionnaires and can assess functional status as well as mental. The SF-36 has become the standard for many reasons including brevity and validity. Patient Satisfaction measures assess the patients satisfaction with the outcome of the procedure. Many types of these questionnaires exist and to this date a standard has not been developed. Seperately, these data collection instruments cannot give an accurate assessment of the outcome. It is well known that a technical success (measured by Clinical measures) does not always correlate to a satisfied patient. Also, a patient may be satisfied with the procedure even though it is a technical failure.

"The Ultimate goal is to find pre-treatment factors that predict success. Yet we must understand what success means in patient terms."

Outcomes is:
(Keller, R.)

1. Analysis of Databases
2. Structured Reviews of the literature- Meta-Analysis
3. Prospective, patient orientated clinical trials
4. Cost-effectiveness Studies
5. Decision Analysis- mathematical modeling of probabilities of various outcomes.

Typical Shortcomings of Outcomes Research Today:
(Fowler et al.)

1. Assessment of Short Term Results Only
2. Reliance on Medical Records or physician ratings to measure results with little or no systematic input from patients.
3. Measurement of a limited range of treatment results, with general measures of function and perceived well-being particularly likely to
be omitted.
4. The samples of patients studied are small and of undocumented representativeness.
5. There are no appropiate comparison or control data against which to assess the results.

"Good" Outcomes Research Will Provide the Following:
(Keller, R.)

1. Information to define the research priorities in the field of orthopaedics (or other fields for that matter).
2. Accurate, high quality process and outcomes data that will be standardized and comparable across particular specialties.
3. Information that will guide orthopaedists (and other physicians) and their patients in making the best choices of treatment.
4. Data and information that will scientifically document the effectiveness of medical care.

Key Methodological Issues:
(Fowler et al.)

1. The best questions to ask patients to assess how they are affected by their treatments must be developed.
2. The comparative advantages of various study designs must be determined.
3. The way data collection decisions, such as mode of data collection and use of proxy respondents, affect study results.