Two Ways to Study Events Over Time
In Observational Studies and Their Limits, you learned that researchers can record existing conditions without assigning treatments. Observational studies can also be described by their direction in time. Some use information about events that have already happened; others begin with subjects now and follow them into the future.
These approaches can help answer questions about health, education, the environment, or community life. For example, researchers might study whether a past pattern of air quality is associated with later asthma diagnoses. They could look back through existing records, or they could enroll subjects and record air quality and health outcomes as time passes. The central distinction is when the relevant information is collected in relation to the study.
A useful way to classify a study is to ask: At the start of the investigation, had the events or outcomes of interest already happened, or will researchers follow subjects to see what happens next? The calendar date alone does not decide the classification. A study conducted this year can be retrospective if it analyzes records from earlier years. A prospective study might begin this year and continue for several years.
Both approaches can be observational. In either one, researchers may measure characteristics or exposures that already exist rather than assigning them. As in Experiments and Why They Can Show Cause, deliberately assigning a treatment makes a study an experiment; simply following people forward does not.
How the Timelines Differ
A retrospective study generally starts with a question about a past period and uses information already available, such as medical files, school attendance logs, or archived environmental measurements. Researchers may also ask subjects to remember past events. The study can be completed relatively quickly if records are accessible, but the information may not have been collected for the researchers’ exact question.
A prospective study begins before the outcomes of interest occur. Researchers identify subjects, define what information to record, and follow the subjects through a specified future period. They can plan how to measure variables and when to check for outcomes. However, they may have to wait a long time, and some subjects may stop participating before follow-up ends.
| Feature | Retrospective study | Prospective study |
|---|---|---|
| Direction in time | Looks back to earlier events or conditions | Follows subjects forward from the study’s start |
| Typical information source | Existing records or recall of past events | Measurements collected during follow-up |
| Possible advantage | Can use information that is already available | Researchers can plan measurements in advance |
| Possible limitation | Past records may be incomplete or inconsistent | Follow-up takes time, and subjects may drop out |
Neither design is automatically better. The choice depends on the question, what information is available, how quickly an answer is needed, and how well the measurements can represent the variables of interest. As in Deciding Which Variables to Measure, researchers should define what they will record and how they will record it.
Worked Example: Looking Back at School Records
Worked Example: Looking Back at School Records
A research team wants to investigate whether absences during the first semester are associated with final course grades among students who graduated last year. The team reviews attendance logs and final-grade records that the school already has. Is this study retrospective or prospective?
Both the absences and final grades occurred before the research team began its investigation. The team is looking back at existing records, so this is a retrospective observational study. The researchers are not assigning students to miss or attend school; they are examining information about conditions that already occurred.
Before comparing the records, the team should check whether the records cover the intended students and period, whether absences and grades were recorded consistently, and whether any records are missing. If the research question concerns all students who graduated last year, records that exclude some graduates could leave the studied group different from the population of interest. Records also may not explain why a student was absent.
Suppose the team finds that students with more recorded absences tended to have lower final grades. A careful conclusion is that, among the graduates whose records were analyzed, a greater number of first-semester absences was associated with lower final grades. The records do not show that absences caused lower grades. Other circumstances could be related to both attendance and grades, and the observational design does not assign attendance conditions.
This example shows why the study’s timeline and the strength of its conclusion are separate issues. Calling a study retrospective describes how the information was gathered; it does not tell us whether the records are complete or whether the observed association has a particular explanation.
Worked Example: Following Subjects Forward
Worked Example: Following Subjects Forward
A team wants to learn whether time spent gardening is associated with reported stress among adults in a town. At the start of spring, the team asks a group of adults how many hours they garden each week. For the next six months, participants complete a monthly survey about their stress. Researchers do not assign anyone to garden. Is this study retrospective or prospective?
The team measures gardening at the start and then follows the participants to collect later stress reports. Since the study records information as future outcomes occur, it is a prospective observational study. It is not an experiment: participants choose whether and how much to garden, and researchers do not assign gardening as a treatment.
The team should specify the population it wants to learn about, explain how participants were selected, and define the variables consistently. For example, “hours of gardening per week” needs a clear reference period, and the stress measure should be asked in the same way each month. The team should also track how many participants complete each survey. If participants with higher stress are less likely to respond, the final data may not represent everyone who began the study.
Suppose participants who reported more gardening at the start also tended to report lower stress during follow-up. The team could describe this as an association between reported gardening time and reported stress among the participants studied. Following subjects forward helps establish that the gardening measurement came before the later stress reports. However, that timing alone does not establish that gardening caused lower stress; the study is observational and does not rule out other explanations.
Choosing Between the Approaches
A retrospective approach may be practical when the events have already happened and useful records exist. It can be especially helpful when waiting for a future outcome would take too long. Its limitations often come from the quality and purpose of the existing information. Records might use changing definitions over time, omit relevant details, or cover only people who used a particular service. If researchers rely on memory, subjects may not recall past events accurately or consistently.
A prospective approach lets researchers plan data collection before the outcomes occur. They can set a schedule, use consistent definitions, and decide what information to record. These advantages do not guarantee accurate data: subjects may report inaccurately, measurements may be unreliable, and people may stop responding. Prospective studies can also require substantial time and resources, especially when outcomes are uncommon or take years to appear.
In either design, researchers should be precise about who was studied and what the results describe. If participants were volunteers from one clinic, for example, the findings do not automatically describe every person in the larger community. As in Defining the Population of Interest, the target population should be clear, and the sample or group actually observed should not be mistaken for that entire population.
Worked Example: Comparing Two Plans for One Question
Worked Example: Comparing Two Plans for One Question
A parks department wants to study whether neighborhood access to tree cover is associated with summer temperatures inside nearby homes. Describe one retrospective plan and one prospective plan, then identify a limitation shared by both.
For a retrospective plan, the team could identify homes with usable temperature records from last summer and compare those records with existing maps showing tree cover around each home. The team would be looking back at measurements already collected, so the study would be retrospective. The records might not include every home, and indoor temperatures may have been measured with different devices or schedules.
For a prospective plan, the team could select participating homes before the next summer, define how tree cover will be measured, and install the same type of temperature sensor in each home. It would then collect temperature measurements during the summer. Because the team follows the homes forward to record future temperatures, this plan would be prospective. It would take time to complete, and some households might withdraw or fail to keep the sensor operating.
Both plans are observational if the team measures existing tree cover and does not assign trees or change the neighborhood environment. Both could identify an association between tree cover and indoor temperature among the homes studied, but neither by itself would prove that tree cover caused the temperature differences. Other features of homes or neighborhoods could also be related to both tree cover and temperature.
Common Mistakes and AP Exam Tips
- Classifying by the date the study is conducted. A study conducted now can be retrospective if it uses past records. Explain whether researchers look back at already existing information or follow subjects forward.
- Assuming prospective means experimental. A prospective study can be observational. State whether researchers assign a treatment; following subjects into the future is not, on its own, treatment assignment.
- Claiming that prospective studies prove cause and effect. Following subjects forward can show that one measured condition came before a later outcome. It does not replace random assignment or rule out other explanations.
- Treating all past information as equally reliable. Identify a specific possible issue, such as incomplete records, inconsistent definitions, or inaccurate recall. Do not claim that every retrospective study has poor data.
- Ignoring what happens during follow-up. In a prospective study, mention possible nonresponse or loss of participants if it could affect whose information is available at the end.
- Describing an association without naming the group or variables. A strong answer states what was associated, for whom, and in which direction, then avoids saying one variable caused the other unless the design supports that claim.
For a full-credit AP response, identify the study design using its timeline and support the classification with a detail from the scenario. For example: “This is prospective because the researchers measure gardening at the start and follow participants to collect later stress reports.” If asked what the results establish, describe the association in context and note that an observational study does not by itself establish causation.
Check Your Understanding
For each situation, identify the study’s time direction and explain one relevant strength or limitation.
- A researcher reviews archived water-quality measurements from five years ago and compares them with records of algae growth from that summer. Is this retrospective or prospective?
- A team measures current exercise habits and surveys participants every month for the next year about sleep. What type of study is this, and why is it not necessarily an experiment?
- A clinic study uses patient records, but some years have no recorded information about a key measurement. What limitation should researchers consider?
- Researchers follow residents into next winter to record home heating use. Some residents stop responding. How might this affect the study?
- Why does following subjects forward not, by itself, prove that an observed exposure caused a later outcome?