Understanding the determine-basal logic

The core, lowest level logic behind any oref0 implementation of OpenAPS can be found in oref0/lib/determine-basal/determine-basal.js. That code pulls together the required inputs (namely, recent CGM readings, current pump settings, including insulin on board and carbohydrates consumed, and your profile settings) and performs the calculations to make the recommended changes in temp basal rates that OpenAPS could/will enact.

Short of reading the actual code, one way to start to understand the key determine-basal logic is to understand the inputs passed into the script/program and how to interpret the outputs from the script/program.

Summary of inputs

The determine-basal algorithm requires 4 input files:

  • iob.json
  • currenttemp.json
  • glucose.json
  • profile.json

In addition, the algorithm can accept 2 optional input files:

  • meal.json
  • autosens.json

When running oref0-determine-basal.js, the first thing you’ll see is a summary of all the inputs, which might look something like this:

{"carbs":0,"boluses":0}
{"delta":5,"glucose":161,"short_avgdelta":4.5,"long_avgdelta":3.92}
{"duration":0,"rate":0,"temp":"absolute"}
{"iob":0,"activity":0,"bolussnooze":0,"basaliob":0,"netbasalinsulin":0,"hightempinsulin":0,"time":"2017-03-17T00:34:51.000Z"}
{"carbs_hr":28,"max_iob":1,"dia":3,"type":"current","current_basal":1.1,"max_daily_basal":1.3,"max_basal":3,"max_bg":120,"min_bg":115,"carbratio":10,"sens":40}
  • meal.json = {"carbs":0,"boluses":0}

    • carbs = # of carbs consumed
    • boluses = amount of bolus insulin delivered

    Those data come from what you entered into your pump or Nightscout web app. If provided, allows determine-basal to decide when it is appropriate to enable Meal Assist.

  • glucose.json = {"delta":5,"glucose":161,"short_avgdelta":4.5,"long_avgdelta":3.92}

    • delta = change in BG between glucose (most recent BG) and an average of BG value from between 2.5 and 7.5 minutes ago (usually just a single BG value from 5 minutes ago)
    • glucose = most recent BG
    • short_avgdelta = change in BG between glucose (most recent BG) and an average of BG values from between 2.5 and 17.5 minutes ago (that average represents what BG levels were approximately 10 minutes ago)
    • long_avgdelta = change in BG between glucose (most recent BG) and an average of BG values from between 17.5 and 42.5 minutes ago (that average represents what BG levels were approximately 30 minutes ago)

    Those data come from your connected CGM or from your Nightscout web app.

  • temp_basal.json = {"duration":0,"rate":0,"temp":"absolute"}

    • duration = Length of time temp basal will run. A duration of 0 indicates none is running.
    • rate = Units/hr basal rate is set to
    • temp = Type of temporary basal rate in use. OpenAPS uses absolute basal rates only.

    Those data come from your pump.

  • iob.json = {"iob":0,"activity":0,"bolussnooze":0,"basaliob":0,"netbasalinsulin":0,"hightempinsulin":0,"time":"2017-03-17T00:34:51.000Z"}

    • iob = Units of Insulin on Board (IOB), net of your pre-programmed basal rates. Net IOB takes all pre-programmed basal, OpenAPS temp basal, and bolus insulin into account. Note: iob can be negative when OpenAPS temp basal rate is below your pre-programmed basal rate (referred to as “low-temping”).
    • activity = Units of insulin active in the previous minute. Approximately equal to (net IOB, 1 minute ago) - (net IOB, now).
    • bolussnooze = Units of bolus IOB, if duration of insulin activity (DIA) was half what you specified in your pump settings. (dia_bolussnooze_divisor in profile.json is set by default to equal 2, but you may adjust this if you’d like OpenAPS to activate a low-temp sooner or later after bolusing.) bolussnooze is used in oref0-determine-basal.js to determine how long to avoid low-temping after a bolus while waiting for carbs to kick in.
    • basaliob = Units of net basal Insulin on Board (IOB). This value does not include the IOB effects of boluses; just the difference between OpenAPS temp basal rates and your pre-programmed basal rates. As such, this value can be negative when OpenAPS has set a low-temp basal rate. Note: max_iob (described below) provides a constraint on how high this value can be. The determine-basal logic will not recommend a temp basal rate that will result in basaliob being greater than max_iob.
    • netbasalinsulin = this variable isn’t used in OpenAPS logic anymore, but hasn’t been removed from iob.json yet.
    • hightempinsulin = this variable isn’t used in OpenAPS logic anymore, but hasn’t been removed from iob.json yet.
    • time = current time

    Those data are calculated based on information received from your pump.

  • preferences.json = {"carbs_hr":28,"max_iob":1,"dia":3,"type":"current","current_basal":1.1,"max_daily_basal":1.3,"max_basal":3,"max_bg":120,"min_bg":115,"carbratio":10,"sens":40}

    • Contains all of the user’s relevant pump settings
    • max_iob = maximum amount of net IOB that OpenAPS will ever allow when setting a high-temp basal rate. This is an important safety measure and integral part of the OpenAPS design. You should set this value based on your current basal rates and insulin sensitivity factor (ISF, or sens in the OpenAPS code) and after studying how the OpenAPS algorithm performs in low-glucose suspend mode for (at least) several days.

    Those data are set during the OpenAPS setup script (or modified by you directly) and based on information received from your pump.

Summary of outputs

After displaying the summary of all input data, oref0-determine-basal outputs a recommended temp basal JSON (stored in suggested.json), which includes an explanation of why it’s recommending that. It might look something like this:

{"temp":"absolute","bg":110,"tick":-2,"eventualBG":95,"snoozeBG":95,"mealAssist":"Off: Carbs: 0 Boluses: 0 Target: 117.5 Deviation: -15 BGI: 0","reason":"Eventual BG 95<115, setting -1.15U/hr","duration":30,"rate":0}

In this case, BG is 110 mg/dL (6.1 mmol/L), and falling slowly. With zero IOB, you would expect BG to be flat, so the falling BG generates a “deviation” from what’s expected. In this case, because avgdelta is -2.5 mg/dL/5m (-0.1 mmol/L/5m), vs. BGI of 0, that avgdelta is extrapolated out for the next 30 minutes, resulting in a deviation of -15 mg/dL (-0.8 mmol/L).

deviation = avgdelta * 6 (or every 5 minutes for the next 30 minutes) = -15 mg/dL (-0.8 mmol/L).

The deviation is then applied to the current BG to get an eventualBG of 95 mg/dL (5.3 mmol/L). There is no bolussnooze IOB, so snoozeBG is also 95 mg/dL (5.3 mmol/L), and because (among other things) avgdelta is negative, mealAssist remains off. To correct from 95 mg/dL (5.3 mmol/L) up to 115 mg/dL (6.4 mmol/L) would require a -1.15U/hr temp for 30m, and since that is impossibly low, determine-basal recommends setting a temp basal to zero and stopping all insulin delivery for now.

Full definition of suggested.json:

  • temp = type of temporary basal - always “absolute”
  • bg = current blood glucose
  • tick = change since last blood glucose
  • eventualBG = predicted value of blood glucose (based on openaps logic) after full effect of IOB
  • snoozeBG = predicted value of blood glucose adjusted for bolussnooze IOB
  • predBGs = predicted blood sugars over next N minutes based on OpenAPS logic, in 5 minute increments
  • IOB = net insulin on board
  • reason = summary of why the decision was made
  • duration = time for recommended temp basal
  • rate = rate of recommended temp basal in units/hour

Understanding “Advanced Meal Assist” (AMA) and the three purple lines

OpenAPS can high-temp more quickly after a meal bolus if it knows about carbs (which you can enter from either the bolus wizard, or if your rig is online, via Nightscout and/or IFTTT (Alexa, Siri, Pebble).

With AMA, once you enable forecast display in your Nightscout configuration, you will be able to see multiple purple line predictions. To do this, click the three dots next to your timeframe horizon (3HR, 6HR, 12HR, 24HR) and then enable “Show OpenAPS Forecasts”. Once enabled, you will have multiple purple line predictions in Nightscout. (Unless you have NO carbs onboard, then you will have only one purple line.)

  • (Usually) Top line == assumes 10 mg/dL/5m carb (0.6 mmol/L/5m) absorption - aka “aCOB” (see notes below about removing it in 0.6.0 and later)
  • (Usually) Middle line == based on current carb absorption (if current carb absorption is > 10 mg/dL/5m, this will end up being the top line)
  • Bottom line == based on insulin only

The purple lines in Nightscout have changed beginning in oref0 0.6.0 and beyond. We are removing aCOB (unused by everyone) and subbing in a “zero temp”-predBG line (ZTpredBG, etc.) showing how long it will take BG to level off at/above target if deviations suddenly cease and we run a zero temp until then. It then uses the lowest of those ZTPredBGs to ensure that we’re dosing safely with UAM: if UAM predicts BG to stay high (based on deviations), but we wouldn’t be able to prevent a low if those deviations suddenly stopped, we’ll adjust the UAM-based insulinReq to be somewhat less aggressive. Conversely, if UAM shows BG ending up below target in 4h, but the ZTPredBGs show that we can safely zero temp and level out well above target, it will allow oref0 to be slightly more aggressive and bring BG down faster, safe in the knowledge it can zero temp if needed. NOTE: you will require an update to NS to view the new line.

Using temporary targets for “Eating Soon” mode and “Activity” modes

Setting temporary targets is a great way to smoothly (safely) make adjustments using your closed loop instead of manually having to make as many adjustments. The two most frequent scenarios for temp targets are when you plan on eating or exercising, and you want to get all of your parameters (e.g., Insulin On Board - IOB) at an ideal value to keep BGs as flat as possible. Temporary or “temp” targets are ideal for when you want the loop to adjust to a different target level for a short period of time (temporarily) and don’t want to hassle with changing targets in the pump (which is where OpenAPS normally pulls targets from).

Let’s take “Eating Soon Mode” as an example. You know you’ll be eating lunch in an hour, so you set a temporary target of 80 mg/dL (4.4 mmol/L) over the next hour. Because your blood glucose will go lower, it’s a good practice to make sure your sensor is reliable and check you blood glucose before setting the temporary target. Otherwise you may cause a hypo. If your OpenAPS rig is connected to the Internet, it will pull the target from Nightscout and treat to the temporary target, rather than your usual closed loop target. The outcome of this means you’ll have more IOB peaking closer to when you’re eating, and that (along with your usual meal bolus) will help reduce the post-meal BG spike. There’s some really important background information you should understand for this, so be sure and read through the DIYPS blog entries about it (How to do “eating soon” mode and Picture this: How to do “eating soon” mode).

“Activity Mode” sets temporary targets that are higher than your normal targets, so you will want to set these with knowledge of how different activity affects your BG. You’re doing essentially the same things as with Eating Soon Mode, but instead of setting a lower target to increase IOB, you’re setting a higher target to provide a “cushion” to avoid a low that may occur as a result of the activity you’re planning. With a higher target, if connected to the Internet so your rig knows about it, it will only do temps with the temporary high target, to provide that “cushion”. You may want to consider setting activity mode prior to activity, in order to reduce the peak net IOB you have when activity commences. (Example would be setting activity mode to 140 mg/dL (7.8 mmol/L) one hour before you go for a hard run after dinner, to help reduce the impact of any meal time insulin that would otherwise be peaking at the time. You may still have to do carbs or other strategies to fully prevent lows, but this is one approach to help - as is being aware of net IOB going into any type of activity.)

Once you have a good idea of how you would set those parameters for your system, you’ll be ready to set some temporary targets. You can test this out manually by entering a temporary target using Care Portal in Nightscout. You’ll see your temporary target appear as a grey bar in Nightscout, spanning the length of the time frame you entered. You can also cancel the temporary targets via Care Portal.

When you’re ready to use this regularly, you can also configure IFTTT to trigger a temporary basal from your Pebble/Apple Watch, etc., or via an Alexa skill. You’ll find a breakdown of how to use IFTTT on the next page.

Exploring further

For each different situation, the determine-basal output will be slightly different, but it should always provide a reasonable recommendation and list any temp basal that would be needed to start bringing BG back to target. If you are unclear on why it is making a particular recommendation, you can explore further by searching lib/determine-basal/determine-basal.js (the library with the core decision tree logic) for the keywords in the reason field (for example, “setting” in this case would find a line (rT.reason += ", setting " + rate + "U/hr";) matching the output above, and from there you could read up and see what if clauses resulted in making that decision. In this case, it was because (working backwards) if (snoozeBG > profile.min_bg) was false (so we took the else), but if (eventualBG < profile.min_bg) was true (with the explanatory comment to tell you that means “if eventual BG is below target”).

If after reading through the code you are still unclear as to why determine-basal made a given decision (or think it may be the wrong decision for the situation), please join the #intend-to-bolus channel on Gitter, paste your output and any other context, and we’ll be happy to discuss with you what it was doing and why, and whether that’s the best thing to do in that and similar situations.

Note about Square Boluses, Dual Wave Boluses, and Basal Pump Settings of Zero

Due to the way the Medtronic pumps operate, temp basals can only be set when there is no bolus running, including extended (square) / dual wave boluses.

Thus, if you use an extended bolus for carb heavy meals (e.g., pizza), which may still be the optimal approach for you, OpenAPS will not be able to provide temp basals during the extended bolus.

If you have periods in the day where your pump normally has basal settings of zero, your loop will not work! You can resolve this by setting the lowest possible basal setting your pump will permit. OpenAPS will then issue temp basals of zero, as needed.