Participant adherence APIs

Initially we referred to this as compliance, but because “compliance” also has meaning in a governance context, we’re using “adherence” to describe the measurement of study member participation in the study.

In the v2 science APIs, Bridge has three APIs that are involved in schedule adherence:

1. The schedule itself, which is finite in length and described by a timeline;

2. The set of adherence records for the participant which describe what the participant has done in that timeline (we only work with the session records when calculating adherence);

3. A set of event timestamps that tell us where the participant is at in the timeline (ie what the user should have done and what they should currently be doing in the schedule).

So far the Bridge server provides this information through separate APIs and does not keep track of the time of the participants. For the sake of efficiency, I think we want the Bridge server to combine this information and provide reports on the status of the account, which will be less resource intensive that serving out all the information to be processed by a client or worker process. The issues I see with this:

1. For the first time the server will need to have knowledge of the participant’s time and time zone(?);

2. Because this information depends on the time of the request, it is not very cacheable;

3. The reports probably update infrequently compared to the amount they will be read (the exact amount depends on many factors). We will have to devise caching strategies to make the system performant when looking at many participants.

Persistent time windows will be excluded from all adherence reports. Completing assessments that are part of a persistent window do not count for or against adherence.

All of these views operate on the most recent timestamps for all events. Building schedules that rely on a mutable event changing, and triggering a new timeline of sessions to perform, will not work with these adherence APIs. That would be events like “do X when session type Y has been completed.” Since it will show compliance with the most event time stream, it may be sufficient anyway. Past time streams are no longer actionable.

APIs

Each report builds on the latter report.

EventStream Report (all adherence data for one participant)

This report is intended to support the following type of adherence chart:

Open

The report is calculated against the caller’s clientTimeZone value (if they have a value set), or else it is calculated against the timeZone value of the Study (to determine the current date). This is what the JSON would like like:

EventStreamWindow includes the state of that session (which is the basis for adherence calculation). This adherence report could be displayed on a calendar since it is specific to a user and the server must know all the calendrical values to calculate it (it's the only report that can take a timestamp to indicate the time at which the report should be calculated, and it is calculated in the user’s local time if possible).

The completion states are:

We can calculate a compliance percentage from these values across a participant’s entire participation in the study. In the weekly report below, we can calculate adherence for that week alone.

Weekly Adherence Report

This report always returns seven days of adherence records for a given user. The report calculates this information by finding the “week since event N” for every event and every session (it will be the week that has the day that falls on “today” as measured in the user’s local time zone). For performance reasons, we will not calculate this week outside of the “current week” so it’s not possible to move forward or backward from this view (at least initially).

Note that the the individual sessions listed in this report are not lined up by calendar date (unless they were triggered by the same event). The structure of this report would be as follows:

This API would calculate and persist this report before returning it in the call, so that this call can be used by a worker process to build a record for every participant. The call could also be used to retrieve up-to-date state information for a specific participant. If this call is not called for a user, it will not show up in the paginated view of these records. We may want to hide a record that hasn’t been updated in the last 24 hours although a worker process will hopefully force a refresh on all accounts.

The SQL table for this report will not attempt to store the report normalized:

The “labels“ columns will pull out the labels from the JSON into a pipe-delimited string, to enable searching for records that match the label in the study-scoped API (searching by label and percentage of adherence are UI design requirements). Note that right now, this would mean the server would need to create the labels and the client would need to search with those server-supplied labels.

Weekly Adherence Reports for Study

This API is intended to support the following kind of adherence overview for the study:

Open

The final API will allow an admin to retrieve a paged set of weekly adherence records for all participants in the study. This retrieves the records from the database and does not calculate them. The paging search criteria should be the same as the study participants search API. In addition to the search arguments in the /v5/studies/{studyId}/participants/search API, it should be possible to filter and or sort on the following fields:

• The adherence percentage;

• The label of individual windows (through a like query on the labels column).

I would like to reuse the search code for participants, but it may be necessary to join tables in a way where that can’t be done, while still maintaining the desired paging logic. To start, it’s probably possible to just search and sort by adherence and labels.

Protocol adherence and notifications (worker process)

Since all the reports give an adherence percentage (whole study or one week), we can use this to sort records by adherence in the weekly view, or to generate a message to administrators as part of the worker process that will be run nightly to update the weekly reports. Any necessary configuration for this can be added to the Study model (in particular, at what percentage of adherence should we notify administrators that a participant is out of adherence).

A worker process could run hourly, querying Bridge for any studies that had a time zone such that the local time of the study was at a specific time (e.g. 4am). If that study had a schedule, all the participants would be retrieved and the weekly adherence report would be generated for them. Since these reports are being generated on the server’s time, it’s possible in unusual scenarios that this time would give odd results for a user in a far off time zone…if so we can possibly run reports for a study more frequently than every 24 hours (but we needn’t start by implementing to prevent this).

If the worker hit certain criteria, it could also send a message warning about a participant being out of adherence.

Adherence Service

The following methods will be added to the AdherenceService:

System Messaging APIs

We have had requirements to message both participants and study administrators over the years. For example, very early on we embedded a blog in an app for this purpose. Then we implemented support for remote push notifications and topics (rarely used). Instead in recent studies we have tried using local notifications based on scheduling information. Now we are talking about showing notifications to administrators about participants being out-of-adherence in an administrative UI. Because “notifications” is heavily overloaded for Bridge, I will call these “system messages.”

I would like to add a simple Ωmessaging system to Bridge which could be used to persist, deliver, and record the consumption of messages. Features would include:

• API would be pub-sub and pull-based (write a message, request a list of active messages for a recipient; push-push via something like SMS or email is a lot more complicated);

• Recipients should include individuals only, opt-in, to start;

• Each recipient could mark the message as “read” to hide it from their UIs, or “resolved” to hide it for everyone. They would not be deleted so they are auditable. We should record who deleted or resolved the message;

• Messages would have a type (“out of adherence,” “study joined,” “study completed”, etc.) and messages could be filtered on the type and/or study;

• These should be app-scoped because it would be very annoying to have to go into every study to see what needed attention in that study.

This should meet the basics needs of our recent UI designs.

The objects (subscriptions and message):

Additional proposed work items

• Allow study burst and custom event IDs to have spaces, essentially allowing them to be human-readable labels like “Study Burst” or “Clinic Visit.” Right now we prevent the obvious labels from being used but this becomes difficult when we’re going to allow people to search by labels that include these identifiers.

• Study should add a studyTimeZone field. We may also want to add an adherenceThresholdPercent in anticipation of messaging.