Week 9. Evaluating Prognostic Literature Flashcards
Define prognosis
examining possible outcomes of a condition or disease and the likelihood they will occur baed on patient “presenting” or baseline characteristics.
Example of patient questions
- when can I return to playing sport
- when will I walk? Will I ever walk again?
- Will I get better?
- When can I go back to work?
Clinical importance of accurate prognosis?
- Expectation:
helps therapist educate patient about appropriate expectations - Defined outcomes:
define expected outcomes from therapy - Prevent bad outcomes:
possibly prevent bad outcomes through education and intervention
Caution with prognosis
we should not divulge prognosis just because we know what they are. Some pt don’t want to know their prognosis, especially if it is not good. Know when and how to inform patient of poor prognosis
What is prognostic factor?
Also known as determinant
A presenting or baseline characteristic that is measurable and associated with patient’s eventual outcome.
Example of prognostic factor (determinant)
- demographics (age, sex)
- disease/condition (e.g. previous injury; pre-injury status)
- Clinical status (severity, level of disability)
- Comorbidity (other medical condition)
It is not an intervention
Why does prognosis precede therapy
prognosis = observation therapy = interventional studies
- identify modifiable characteristics through observation
- test what happens when characteristics are modified through intervention
Prognostic factors: modifiable vs. non-modifiable
Non-mod.
- age
- sex
- social-economic
- culture
- past history
- genetics
- comorbidity
Mod.
- pre-condition status
- strength
- ROM
- balance
- belief, emotion, behaviour
- loading demand
- lifestyle
Potential Outcome Measures
- Survival
- patient, implant, repair - Recovery
- rate and time - Recurrence/re-injury
Prognostic factor vs. risk factor
RISK: associated with development of a disease/condition (before onset)
PROGNOSTIC: associated with recovery form a disease/condition (after onset)
risk –> risk factor –> condition/event –> prognostic factor –> prognosis
Prognostic Study designs: RTC
cannot do RCT on prognosis because cannot assign or randomize subjects, but data from RCT are used in study prognosis
Prognostic Study design: observation
- PS are observational because cannot assign or randomize subjects, rather just have to look at different characteristics.
- look at outcomes associated with having or not having a factor at time of study entry
Prognostic Study Design: level 1 evidence for study
Prospective cohort (follow patients over time) = Longitudinal study
- Natural course = follow untreated patients
- Clinical course = follow patients treated in usual way
Longitudinal Design
A study of the same participant(s) at more than one point in time
- Prospective (enrol –> follow up)
- Retrospective (enrolled –> recall)
Inception cohort Design
Patients at similar point in their condition
- Incident = new cases (best for prognosis studies)
- Prevalent = any cases (including pts who already have bad outcomes, survivor cohorts)
Levels of Prognostic Study Designs
Level 1: SR of prospective cohort studies
Level 2: Prospective/inception cohort study
Level 3-4: dependent upon quality
- retrospective cohort
- case-control, case series
Example of level of evidence: “What will happen if we do not add a therapy”
Level 1: Systematic review of inception cohort studies
Level 2: inception cohort studies
Level 3: Cohort study or control arm of randomized trial
Level 4: case-series or case-control studies, or poor quality prognostic cohort studies
What is the main factor that must be included in prognoses
Prognosis have to be associated with a time
e. g.
1) 80% chance of pt with grade 2 lateral ankle sprain returning to sport within 1 month
2) 11% 5-year risk of injury contralateral ACL in patient undergoing ACL reconstruction
3) 20% 10-year risk of LT work disability in patients with MS
Search terms:
‘predict’
‘prognos*’
‘prognostic factor’
‘predictive validity’
Building Prognosis Question: what do we need to consider
- Population/patient
- how do I describe patient similar to mine - prognostic factor
- which predictive factors I using - outcome/timeframe
- what disease progression can be expected?
~38 year old female with history of right shoulder rotator cuff repair
Post-surgical protocol said no skiing or heavy lifting for 1 year
After 1 year the patient wants to return to skiing, but is afraid of re-injury
Asking whether it is ok to return to ski
Research question?
Does preoperative quads strength affect postoperative function after Total Knee Arthroplasty?
How do we know if question is prognosis or intervention study?
Can we randomize population group? No, then likely a prognosis.
Example of prognosis vs. intervention question for TKA.
Prognosis
Does preoperative quads strength affect postoperative function after Total Knee Arthroplasty?
Intervention
Does quads strengthening prior to Total Knee Arthroplasty improve function?
Evaluate using critical appraisal tool: overarching/essential questions
- Are the Results Valid? (Review methods)
- What are the Results? (Review & interpret the results)
- Can I apply them to my patient(s)? (Look at inclusion/exclusion for applicability, quality of study, change in outcome associated with a factor)
Prognosis Study Appraisal is similar to?
appraisal is very similar to appraisal of RCTs
Criteria for determining if the results are valid
1) Was there representative sampling from a well-defined population?
2) Was there an inception cohort?
3) Was there complete or near complete follow-up?
4) Were objective, unbiased outcome criteria used?
5) Was there adjustment for important prognostic factors and potential cofounder?
Was there representative sampling from a well-defined population?
- start by looking whether the study recruited “all” patient or “consecutive” cases
If not, study may provided biased estimates of true prognosis
Was there an inception cohort?
similar point in their condition
- Incident = new cases (best for prognosis)
- Prevalent = any cases (includes patient who may already have bad outcomes, survivor cohort)
Follow-up
Follow-up long enough for events to occur
- completion rate should be >80% (ideally >85%)
Were objective, unbiased outcome criteria used?
- measurable
- reproducible
Adjustments for prognostic factors and potential confounders?
confounder = variable that correlates with both the dependent and independent variables in a way that explains a way some or all of the correlation between these two variable
Example: 3rd factor (being Z) is the potential confounder
Every time Z changes X and Y have the potential to change
If X changes due to this and Z is ignored it would be assumed Y is the changer when really Z is the confounder
Confounding example: does age affect pain after TKA
- outcome
- prognostic factor
- confounder
- outcome = pain
- prognostic factor = age
- confounder = type of implant used (affects pain, implant closed based upon age)
Example: Recovery of function after thoracic surgery
- outcome
- prognostic factor
- confounder
- outcome: function
- prognostic factor: pre-operative health status (age, pain intensity, severity of condition, supportive family)
- confounder: comorbidity
Adjustments for all important prognostic factors/potential confounders
Sample size:
- typically need 10 patients that had the outcome for every 1 factor in analysis
- large samples can use a lot of potential factors
How likely are the outcomes: mean difference
compare mean differences (e.g. function) between two groups
How likely are the outcomes: odds/risks ratio
predict odds or risk of events occurring in the presence of prognostic factors in a single point in time
How likely are the outcomes: hazard ratios
predict odds of event occurring in the presence/absence of prognostic factors over time
How likely are the outcomes: Survival curve
Graph of discrete events/non-events occurring over a defined time. In most clinical situations, the chance of an outcome changes with time. Earlier follow-up results are more precise because there are more patients.
Diagnostic 2x2 table a-d
a = true + b = false + c = false - d = false +
Formula
odds ratio
risk ratio
odds ration = ad/bc
risk ratio = [a/(a+b)] / [c/(c+d)]
Written explanation of OR and RR
OR = # times the ODDS more likely to
RR = # times more likely to
Interpreting statistics: Odds/Risk/Hazard Ratio
Odds/Risk/Hazard Ratio
= 1 no increased/decreased risk of the outcome occurring associated with that factor
> 1 = outcome expect to occur more often in exposed relative to non-exposed
< 1 = outcome expected to occur less in exposed relative to non-exposed
Why do we use 95% CI to determine how precise are estimates of likelihood?
1) CI indicate strength of effect and precision of estimates
2) includes estimate of effect and range (interval) of possible estimate if study were repeated
3) Wide CI = less precise estimate of effect
4) Narrow CI = precise estimate of effect
How to narrow CI
- increase sample size
- increase number or outcomes
What does precision indicate?
does not give magnitude of effect, but rather the possible range for magnitude
- precise & significant
- precise & NS
- non-precise & S
- non-precise & NS
Adjustment of 2x2 table: univariable
considers 1 factor and its association with an outcome (aka univariable or invariable analysis)
Adjustment of 2x2 table: multivariable
more commonly, we look at impact of multiple factors and their association with the outcome at the same time (aka multivariable analysis)
- looks at the independent effect or a factor after controlling or adjusting for the other factors
Clinical significance: patient type
were study patients similar to mine (examine inclusion and exclusion criteria)
Clinical significance: patient experience
will results help me re-assure/inform patients? (does it change what I would tell patients to expect?)
Why do we use prognosis?
- education patients about appropriate expectations
- define expected outcome from therapy
- possible prevent bad outcomes through education/intervention
