A lot of medical tests (medical test (Q2671652)) are conducted on a daily basis, by physicians or other members of medical staff. Many of the tests have reference range (Q1626599), which is used to interpret the results of the test, but to my knowledge, no other open database provides structured data of reference values. This property will be useful to make a diagnosis support system.
Some reference values differ with genders, age, countries, or other factors. If the upper limit of the reference value and lower limit of it are separately stored, it is very inconvenient to use and to maintain. So I propose we adopt "dummy items and qualifiers" solution for this property, like the way used in union of (P2737). Okkn (talk) 12:14, 27 February 2018 (UTC)
Comment@Okkn: Please provide an example of an item data type on the first property.Thanks David (talk) 15:12, 27 February 2018 (UTC)
@ديفيد عادل وهبة خليل 2: Thank you for your comment. The subject of the first property is an instance of reference range (Q1626599), such as items of blood tests, blood pressure, respiratory function test, and so on. And the data type of the value of the first property is item, and allowed values are dummy items named something like "reference range as qualifier", "one-sided cutoff value as qualifier", and "equivocal interval as qualifier", but I haven't created them. --Okkn (talk) 17:12, 27 February 2018 (UTC)
@ديفيد عادل وهبة خليل 2: Let me know if you have still any questions or need clarifications on these properties. Many thanks, --Okkn (talk) 05:58, 10 March 2018 (UTC)
Not done Lack of support. Micru (talk) 08:53, 11 April 2018 (UTC)
Although the discussion is stagnant, no one oppose this proposal. @Was a bee: If it's not too much trouble, could you please give me some comments? --Okkn (talk) 11:22, 11 April 2018 (UTC)
Comment @Okkn: This is interesting property. If such data exist in Wikidata, that's nice. But at the same time, I feel one difficulty. "Normal range" can vary depend on source by source. Although almost same range are shared by all sources, exact threshold would be blur. I have read that in medicine, sometimes outside of 2 standard deviation (2σ) is treated as abnormal. But average and distribution varies depend on samples (age, sex, race, etc). And pessimicstic docters/researchers would take narrower normal range, and optimistic take broader. How can we think about this kind of blurriness?--Was a bee (talk) 20:14, 12 April 2018 (UTC)
@Okkn: OK. But it is difficult for me to imagine actual usage. So I made example with Template:Statement+. Which has 2 values, one for adult and one for child (values and references are randomly pick-uped. It doesn't have meaning). Is this usage right? Or if wrong, please fix that. --Was a bee (talk) 02:22, 14 April 2018 (UTC)
Compared with these property's name, "lower limit" and "upper limit" is very broader name (in other words, it seems that these properties looks usable at everywhere). So I think that it is better to...
1. Change description of "lower limit" and "upper limit" into "less restrictive" text. For example, change to "You can use this property everywhere".
2. Or change property name to more restrictive name. For example, change to "upper limit (reference value)" and "lower limit (reference value)"
@Was a bee: You are right. I changed the descriptions of the qualifiers to less restrictive texts. In the future if someone create a property that adopts the dummy value solution to store the range of values, these general qualifiers will be useful. Also, if the subjective item is about an interval, such as closed interval from −1 to 1 (Q21813730), I think these properties can be used as usual properties. --Okkn (talk) 10:51, 15 April 2018 (UTC)
@Okkn: OK. Then finally let me make sure. If properties are created, do you import certain amount of data? Because this is interesting but complex data. So if you don't import, I think no one won't import. Formerly, you imported very complex thousand of data, after properties which you proposed were created (Wikidata:Property proposal/recognition sequence). So I assume the same in this case. Am I right? --Was a bee (talk) 19:45, 16 April 2018 (UTC)
@Was a bee: Of course! I will use this property effectively. At first, I will import all of the JCCLS Common Reference Intervals (PDF) to show how to use this property. --Okkn (talk) 00:46, 17 April 2018 (UTC)
Thank you. Then all my questions have been solved. So I put here icon Support. --Was a bee (talk) 20:15, 17 April 2018 (UTC)
@Was a bee: I appreciate your kind advice and support! --Okkn (talk) 05:51, 18 April 2018 (UTC)
@Okkn: I'm a bit torn about the use of dummy items for values - I think we should generally try to avoid that as much as we can. Did you know that quantities in Wikidata can have lower and upper bounds? This can be inserted in the UI by typing quantities such as "1,034.56±3.6". Don't you think that it would be more suitable? − Pintoch (talk) 07:57, 28 April 2018 (UTC)
@Pintoch: I know it, but I gave up on the idea of using the lower and upper bounds of quantity data type for two reasons. First, not all reference values have an ideal value and some are not normally distributed, so we can't express them as "true value ± tolerance" or "mean ± 2SD". Second, some medical tests have only one-sided cutoff value, so we can't use the native data type. The advantages of using dummy items include that we can store all types of reference values in one way. Does my answer make sense? --Okkn (talk) 09:51, 28 April 2018 (UTC)
Yes it makes sense, especially given the fact that it is not possible to give only one upper bound for instance (so your examples where there is only one bound could not be modeled in this way). Given that I am not entirely sure about this, and considering that only few people have commented here so far, I will hold off from creating this right now, just in case someone comes up with a better solution soon. − Pintoch (talk) 12:06, 28 April 2018 (UTC)
When I look at the values I think the miss a specification about the authority that defined the reference values. I think it would be great if we get clear about how the authority should be specificed before we create this property. ChristianKl ❪✉❫ 17:32, 6 May 2018 (UTC)
In considering the domain of these tests, should it not be made explicit the species the ranges apply to? The same test may be useful for humans, mice, or even C. elegans, but the expected range of results will generally differ.LeadSongDog (talk) 14:27, 7 May 2018 (UTC)
@ChristianKl: It has passed two months since you pinged WikiProjecct Medicine, but there is no better ideas or opinions. I am convinced that this property is necessary for modeling medical attributes and medical findings, which are linked using evaluation of (P5134) and has evaluation (P5133). Could you please support this proposal or create it? Regards, --Okkn (talk) 22:04, 7 July 2018 (UTC)