1) Who is the patient:: In particular what are the demographic characteristics of the patient, for example a 65 year old female with diabetes and hypertension
2) What is the disease or event of interest: For example syncope
3) What is the outcome of interest:  For example, recurrent syncope or cardiac death

Medline  is the best source of information about prognosis, secondary sources are particularly weak here. Often, the control group from a large, RCT or cohort study will provide the best available evidence.   Online databases of Mortality data are available through FedStats and CDCWonder. 

• Search for the MESH term:   exp prognosis.   This is a simple  search filter for prognosis.
• Search the disease
• Combine 1 AND 2
• View an example of this search (under construction)

• Inception cohort: 

Did the authors include patients who were at a similar, early point in their disease, and did they describe the selection process? If the cohort is not marching together in step from the problems inception, its conclusions are suspect.  Retrospective assessments of prognosis are problematic. 
 
• Was follow-up long enough?

Many prognostic studies enroll patients when they experience an index event and follow them up for a few years, the duration of a typical study budget. However, long term (many years) complications may in fact have the greatest overall impact on mortality and morbidity.
 
• Was the outcome determined in a blinded fashion?

The lack of blinding has sunk more than one follow-up study. The evaluator must not know which exposure category the patient falls in, though this may be easier said than done for "soft" outcomes, especially when certain risk factors cannot be masked, such as gender.
 
• Was there adjustment for other important prognostic factors? 

The author must account for or statistically correct for the main determinants of prognosis besides that under study.
 

• What is the prognosis? In other words, how likely is the event to occur over time?  There are several common methods for expressing prognosis:

• PEER: Patient Expected Event Rate (Fractional event free Survival) is the percentage of the inception cohort alive (or without the target outcome) after a specified time period, for example 15% 10 year survival. 
• Median (event free) Survival is the time at which half of the patients have survived (or not incurred the target outcome).
• Average Annual Mortality is the average percent of the cohort  reaching the outcome of interest (death) each year. 
• Life expectancy is the average future lifetime of a person at a specified age (sox)
• Survival curves 

• Were the study patients similar to my own?  If the study patients were not like this patient, the estimated mortality may be an over or under-estimate. 

• Will the result lead directly to avoiding or selecting therapy? To answer this question you must know your patients risk of disease (prognosis), the potential benefit of treatment (Number Needed to Treat, NNT) and the potential Harm of treatment (NNH).   With this knowledge you can make an informed decision concerning treatment. 

• Are the results useful for counseling or reassuring patients? There may be little value in knowing a patient’s prognosis when it cannot lead to a change in care. However, we often minimize the impact of prognostic information on the patient’s emotional well-being, or the family’s anxiety about what the future holds. Not only does a prognosis allow a patient and a family to get their affairs in order, many are reassured when the prognosis is excellent. Both of these results will improve communication, thus might avoid fruitless medical encounters in the future, even when the prognostic information is collected at some cost.