Is it possible to generate only "patients"?

[NeuroinformaticaFBF]

Dear Synthea team

For a project I am using you tool to generate large populations of patient (i.e., persons with a specific condition). I do not need healthy controls. What I am doing now is generating a VERY large population, both with patients and healthy controls, and filtering them while keeping only patients. However, this can become inefficient for diseases with low prevalence, if you need a very large patient population.
Is there a way, like a flag to use in CLI, to generate only patients?

Another question: I would like to create a new disease module, but I struggle to understand where "observation.csv" file is created. I admit I did not read the wiki to the last details, is it explained there? Or, is it there a place where I can find this piece of information?

Many thanks

[jawalonoski]

You can use a "keep module" to specify which patients you want to keep. You can specify the "keep module" on the command line. See https://github.com/synthetichealth/synthea/wiki/Keep-Patients-Module

The observations.csv file is only created if you enable the CSV export. The file will be in the ./output/csv/ folder unless you change the configuration to point elsewhere. See https://github.com/synthetichealth/synthea/wiki/CSV-File-Data-Dictionary

[NeuroinformaticaFBF]

Dear @jawalonoski

Thanks for your answer. However, I am getting some trouble using correctly the keep module.
For example, I created a keep_hypertension.json module, starting from your examples:

{ "name": "keep", "remarks": [ "A blank module" ], "states": { "Initial": { "type": "Initial", "conditional_transition": [ { "transition": "Keep", "condition": { "condition_type": "Or", "conditions": [ { "condition_type": "Active Condition", "codes": [ { "system": "SNOMED-CT", "code": 59621000, "display": "Hypertension" } ] } ] } }, { "transition": "Terminal" } ] }, "Terminal": { "type": "Terminal" }, "Keep": { "type": "Terminal" } }, "gmf_version": 2 }

I then run this command:
./run_synthea -m hypertension -p 1000 Lombardia --keep=/src/main/resources/modules/keep_hypertension.json --exporter.csv.export=true --exporter.baseDirectory="path/to/data/hypertension_ONLY_PZs"

What I expected was to generate 1000 patients, meaning that every subject in my patients.csv file has a corresponding row in conditions.csv file, meaning that all my subjects are patients, thus presenting the condition "hypertension".
Instead, I "only" got ~200 patients in the conditions.csv file.

Is it possible to use the keep module to do what I expect to do? And if the answer is yes, what am I doing wrong?

Many thanks as always

[jawalonoski]

Your keep module looks fine.

Try again, but removing the -m switch, which tends to mess things up. See the warnings on https://github.com/synthetichealth/synthea/wiki/The--M-Feature.

[NeuroinformaticaFBF]

Thanks for your reply.

I tried to run the command without the -m switch and with the same keep file of above: ./run_synthea -p 1000 Lombardia --keep=/src/main/resources/modules/keep_hypertension.json --exporter.csv.export=true --exporter.baseDirectory="path/to/data/hypertension_ONLY_PZs"
This time, it generated ~1000 subjects, and every one of them has a condition. However, still the number of patients with hypertension was ~200. Moreover, almost every patient has several disorder/finding/situation.
Is there a way to make that every patient has hypertension, and ONLY hypertension, with the keep command? I guess a workaround could be to manually tweak the prevalence, I wonder if this can be done by the keep module.

Another thing (tell me if you want to move to another topic). I read the warnings on the M feature. Our goal is to generate large population of subjects with specific conditions, and only that specific condition. So, for now, we used the -m flag basically everywhere. I see now that this is not the intended workflow, and it may be removed in the future.
So I ask you, what is the intended procedure to generate subjects that present a specific disease (for example, hypertension)?

Thanks for your time and patience

[jawalonoski]

The keep switch does not filter out other diseases, it just specifies what you keep.... I suppose you can create logic in the keep module that explicitly excludes patients if they have any other ActiveConditions... but that would be a lot of work to list every disease other than Hypertension.

I'm not sure why your patients don't all have hypertension, without examining everything deeply. They probably do, it just doesn't show up in the data. It could be related to the "years of history" setting (search issues and discussions here)... or maybe something else.

The goal is to create patients. Patients typically have comorbidities. The world is like that, especially people with hypertension. If you insist on patients with only a single ailment, you can post-process the data (i.e., remove the other data) or override the probabilities (i.e., make the incidence or prevalence rates of non-hypertension diseases zero). See https://github.com/synthetichealth/synthea/wiki/Module-Overrides

[aboccali]

Hi Synthea Team,

Congratulations on the great tool you’ve created!

I have a question about the logic behind patient generation. I understand how an individual disease module works — for example, I’ve carefully studied the dementia and hypertension modules — but I’m not sure how the system decides which subjects enter which pathology modules.

Is the assignment to modules based on randomness, or do all subjects created go through all the disease modules? And in that case, how are healthy subjects generated?

Thank you very much for your help!

[jawalonoski]

Is the assignment to modules based on randomness, or do all subjects created go through all the disease modules?

All patients execute all the modules.

And in that case, how are healthy subjects generated?

Randomness, typically associated with the distributed transitions, determines which pathways patients follow, and if they are afflicted or diagnosed with any given disease.

[aboccali]

Thank you!

I have another question: which module assigns blood pressure values to a patient?
I have analyzed the hypertension, vitals, and wellness_encounters modules, but I can’t find where the actual numeric BP values are generated. However, in the observations.csv file, the patients with hypertension do have blood pressure values assigned.

Could you please clarify which module or mechanism is responsible for generating those values?

Thank you again for your help!

[jawalonoski]

See BloodPressureValueGenerator.java and htn_drugs.csv and

synthea/src/main/resources/biometrics.yml

Lines 13 to 20 in 0091254

	metabolic:
	blood_pressure:
	normal:
	systolic: [100,139] # mmHg
	diastolic: [70,89] # mmHg
	hypertensive:
	systolic: [140,160,200] # mmHg
	diastolic: [90,120] # mmHg

[aboccali]

Thank you!

One last question: is it possible to prevent the creation of certain observations, such as QALY, DALY, or any others considered not relevant for a specific project, in order to reduce the size of the observations.csv file?

Thank you again for your support!

[jawalonoski]

There is no configuration setting to disable QALY and DALY. You can manually alter the code to prevent them from being written. You can probably do that by commenting out the lines of code linked below.

Otherwise, the overall philosophy of the project remains the same -- Synthea generates all the relevant data, and then the user post-processes it to filter out what they don't need. So, either delete the observations you don't want, or do not import them into your database, or filter them in python, or whatever applies to your situation.

synthea/src/main/java/org/mitre/synthea/export/CSVExporter.java

Lines 279 to 302 in 0091254

	String[] gbdMetrics = { QualityOfLifeModule.QALY, QualityOfLifeModule.DALY,
	QualityOfLifeModule.QOLS };
	String unit = null;
	for (String score : gbdMetrics) {
	if (score.equals(QualityOfLifeModule.QOLS)) {
	unit = "{score}";
	} else {
	// years in UCUM is "a" for Latin "Annus"
	unit = "a";
	}
	@SuppressWarnings("unchecked")
	Map<Integer, Double> scores = (Map<Integer, Double>) person.attributes.get(score);
	for (Integer year : scores.keySet()) {
	birthDay.set(Calendar.YEAR, year);
	if (birthDay.after(cutOff) && birthDay.before(now)) {
	Observation obs = person.record.new Observation(
	birthDay.getTimeInMillis(), score, scores.get(year));
	obs.unit = unit;
	Code code = new Code("GBD", score, score);
	obs.codes.add(code);
	exportObservation(personID, "", obs);
	}
	}
	}

[aboccali]

Perfect, thank you!!