What is provisional data?
Provisional data refers to any data received and used before it has undergone the full data processing and quality activities that prepare it for typical full reporting use. Since data collection, submission and quality activities are ongoing, provisional data needs to be interpreted with caution.
Data cycles
The nature of provisional data varies depending on how a database operates. Most operate on an annual cycle and have official data submission deadlines. Any data received before the completion of processing, including data quality assessment, is provisional. For databases that receive ongoing data submissions throughout the year (e.g., monthly, quarterly), using provisional data can be much more timely, especially earlier in the reporting year.
At the end of a data cycle, some databases have a hard closure and do not accept any more changes. This applies to the Discharge Abstract Database (DAD) and the National Ambulatory Care Reporting System (NACRS), which hold hospital data. Other databases are always open but still have submission deadlines, after which end-of-cycle processing takes place and changes are expected to be minimal. Examples include the Continuing Care Reporting System (CCRS) and Home Care Reporting System (HCRS), which hold specialized care data.
- You can find more information about CIHI’s databases on the Data Holdings page.
- More information about data submissions, including deadlines, is on the Submit Data and View Standards page.
What you need to know about using provisional data
Provisional data can change
Provisional data for the same population and time period can change as often as every month. Data might change if routine data quality checks uncover errors and data providers correct and resubmit data. It might also change if initial submissions include only partial data that is completed through later in-year submissions. Changes can affect analysis, benchmarks and trends.
Provisional data can be incomplete
Provisional data is more timely than closed-year data, but it may be less complete and/or have other quality issues, such as uneven coverage. By understanding this quality trade-off, you’ll be able to decide whether the data is fit for your intended use. You should also consider these factors:
- Although data submission tends to follow a predictable cycle, data providers may have different submission schedules. For example, some jurisdictions may have mandated monthly submission while others ramp up more gradually over the course of the year.
- Data providers may follow different quality assurance processes that affect how and when they revise or correct their data. For example, some may submit default or unknown data values that are updated later in the year when more definitive information becomes available.
- Time of year, or seasonality, may affect the content of data submissions and subsequent analysis and interpretation. For example, the volume of elective surgeries is generally lower in the summer because of vacations.
- Health system events, disruptions or trends can affect data availability and comparability. For example, pandemics such as COVID-19 affect the entire health system in different ways, and we can expect to see real changes in our data as a result (e.g., decreased emergency department and doctor visits). Smaller events will have impacts at a more localized level (e.g., floods, wildfires). These impacts can include
- Delayed or incomplete data submission from areas under pressure
- Heightened need for data for decision-making that may temporarily affect timeliness and availability (e.g., more frequent or mandated submission)
- Temporary redeployment of resources and/or facilities that change existing data streams
- Introduction of new data elements that may evolve over time
Analysis and reporting
Before conducting analysis using provisional data, consider its coverage and completeness. It is important to understand how well provisional data captures your population of interest. For example, it may be possible to describe an individual region within a province, but not the entire province. At the record level, check the availability of analytical variables by looking at missing, default or unknown value rates. Looking at historical data volumes or trends can help you assess completeness, but you should take health system events into account when doing this.
When reporting on provisional data, consider the intended audience. It is important to clearly identify when data is provisional, and to describe real or possible data quality limitations for your users. This transparency will help ensure they interpret the information appropriately. The following uses of provisional data may not be appropriate:
- Analyses using low volumes or measures that require risk adjustments
- Comparative reporting
- Performance measurement and indicator reporting