Cohors, cohortis (f): a tenth of a legion

Dimensions and definitions

A cohort study follows a group of people from one point in time to another and observes changes that occur during that period. The study can be retrospective, for example reviewing all cases of breast cancer, or prospective, identifying a group of health people or patients and following them through time. Cohort studies may use routine data or data specially collected for the purpose of the study or both. This type of research is useful for studying:-

The outcome of treatment where a randomised controlled trial is impossible, for example the outcome of intensive care or prostatectomy. The findings of this type of study allow quality standards, for example the readmission rate, to be evidence based.
Different approaches to service delivery and management where these cannot be tested by a randomised controlled trial either because the number of units are too small to give a trial of adequate power or because the health service policy makers or managers will not allow their service to be included in such a trial. In addition, cohort studies are also useful means of studying "natural experiments" i.e. when changes are made in the organisation or delivery for political or managerial reasons or where different patterns of care exist in similar settings as a result of history and tradition.

Cohort studies have been used to study:

The effects of staffing changes, for example the effect of introducing on-site physician staffing to intensive care units in hospitals other than teaching centres.
The relationship between volume and quality.
The relationship between status and organisation of the hospital and outcome, for example whether or not it is a teaching or non-teaching hospital or whether the hospital is public or private.
The relationship between the organisation of a clinical service such as an intensive care unit or an emergency.
The relationship between different levels of qualification and outcome.

The balance between direct experimentation, though trials, and observational studies such as breast cancer in Scotland has to be continually reviewed.

It is possible to organise randomised controlled trials to assess the benefits of different types of service or even the method of health care financing. However, for many questions about the relationship between the funding and organisation of care and patient outcomes, well conducted cohort study is the best appropriate form of research design provided that the study is carefully appraised.

Questions to ask about a cohort study

The following are the questions to ask. The answers for the west of Scotland breast cancer study are given after each question:

Is clear information given about the way in which the cohort was recruited?: Very clear - all identified patients, using histological definitions
Were any factors that could have included or excluded more severe cases considered?: No - about 98% of identified patients were traced and their history recorded.
If mortality is an outcome, what steps have been taken to ensure that all deaths have been identified?: Through extensive searching of hospital records and death registrations.
If other measures are used, have the instruments used for measurement been validated?: Does not apply
Was the severity of disease taken into account in the analysis?: Yes: confounding issues like severity of breast cancer were analysed for. This made the result more positive.
Was the presence of other diseases (co-morbidity) taken into account in the analysis?: No. May not apply.

Features to note

When appraising the recruitment of individuals into a cohort study the most important feature to note is the completeness of recruitment. All the subjects in a defined time period should be recruited and any sampling procedure, for example the recruitment of patients admitted on weekdays or between 0900 and 1700 hours, should create suspicion that the study is biased. It can also be useful to ask, "Where did patients go who were not admitted?", because the hospital may have referred the more difficult cases elsewhere or those referrers may have selectively referred only the mild cases to the hospital being studied.

Secondly, it is important to ensure that valid criteria are used. Inpatient mortality is an invalid criterion of the quality of the hospital care because of variations in duration of stay and it is better to use a criterion such as the thirty day or sixty day mortality. When criteria other than morbidity are used, the tools used to measure variables such as pain or quality of life should be validated measuring instruments.

Finally, in the analysis the severity of illness should always be taken into account and therefore explicitly mentioned in the text. In studies of intensive care for example, a validated system for assessing severity - the APACHE (Acute Physiology and Chronic Health Evaluation) system is used. It is also essential to take into account the presence or absence of other diseases which might have influenced outcome, what is called the co-morbidity. Robust techniques have been developed for doing this and must be used if valid results are to be obtained.

Uses and Abuses

The main abuse of a cohort study is as a means of assessing the effectiveness of a particular intervention when the more appropriate method could be a randomised controlled trial. Cohort studies are appropriate when assessing changes in service management or organisation or in searching for uncommon side effects of treatment or the adverse effects of treatment.

previous or next story in this issue