Instead of spending money on training local dentists, it might be more worthwhile to do disruptive innovation. There’s no reason to believe that the eyes of dentists are superior at diagnosis to a well-trained machine learning model.
Dental health care as currently practiced needs a lot of trust. I have to trust my local dentist with their assessment of whether my teeth need costly procedures. Having well-justified trust is easier in the first-world even when first world dentists still differ a lot in their treatment recommendations. In the third world it’s more likely that dentists will abuse trust by selling their customers unnecessary treatments.
If the diagnosis and treatment recommendations would be made by a branded computer program, it would get a lot easier to trust the model. Training costs would also go down if dentists wouldn’t need to be trained in diagnosis anymore.
Interventions could also plausibly be automated but that might be more expensive than just automating the diagnosis.
Hey Christian, thanks for your comment! I totally agree that ML has great potential for diagnosis (in dentistry but also within the field of medical care more broadly– e.g. I was reading about this grant from Gates the other day, it’s a diagnostic ultrasound for maternal conditions). Caveat that I’m not sure that the average person would trust an ML diagnosis over a ‘real person’ diagnosis (at least, not yet), and I think it would be a while until roll-out in an LMIC type context. Nonetheless, I think this is a promising area which I want to check out in more detail.
Within the field of oral health though, I think that investing in preventative treatment is currently more promising than diagnostics/ curative treatment: i.e. while I think that training up local dentists may be cost-effective, it seems a little borderline. However, preventative treatment (aka fluoride) seems to be extremely cost-effective, and we already know how to do it. This is mainly because salt and milk fluoridation appear to work and are insanely cheap—that’s where my $2 WELLBY/ $9 DALY figures are coming from.
Teeth need fluoride but it’s likely harmful to the rest of the body. Application via toothpaste a lot more of the fluoride actually goes to the teeth has advantages over adding it to salt.
It might be that salt and milk fluoridation is still net-beneficial but that calculation is more complicated than just comparing the monetary cost.
As far as trust is concerned plenty of people also don’t trust water, salt or milk fluoridization and programs by Westerns in Africa might face political problems.
Thanks for this comment, I am happy someone drew attention to this!
I did check out various papers (such as the one you linked, and some other papers proposing a link between fluoride exposure and adhd). I didn’t find them convincing enough to add in a non-monetary cost to fluoride (bear in mind that I also left out various potential net-benefits, such as increased attendance at school and work).
To give a bit more detail (because I think this important), I found that the evidence linking fluoride to neurological outcomes was pretty low quality. I could only mainly find ecological studies, where i judge the risk of confounds to be very high- plus other issues that make me suspicious such as exclusion of outliers (without clear justification). That being said, I do think it’s important to be cognisant of these theories, even if the available evidence is poor— I quite like this review, which basically argues that low to adequate levels of fluoride don’t appear to be associated with any neurological conditions, but there is ‘low level’ evidence that they are at high concentrations.
To me, this gets at the point that fluoride should only be used in communities that have low levels of fluoride already— apart the speculation about neurological disorders, we know that fluoride can cause fluorosis (weakening of tooth enamel) in levels with high levels of fluoride. I have put the ‘costs’ of fluoride at zero, but have highlighted in the report that I think these kinds of interventions should only be used in area with low fluoride (for instance, not areas with high levels of groundwater fluoride). In areas with high fluoride, I agree with you that there might be ‘costs’ to be accounted for .
Totally agree with you on the point about trust! I quite like the salt/ milk interventions since they are more ‘opt in’ than water fluoridation (aka you could just buy the other salt/ milk), as well as being more practical in areas with lower levels of piped water access. I’d hope that this would mean less pushback.
The general way to account for side effects of clinical interventions is that low-level evidence is enough to take side effects seriously. Low-level evidence is enough for the FDA to require drug manufacturers to put a warning label on a drug.
The review you point to says “indicating the need for new epidemiological studies that could provide further evidences regarding this possible association”. That means there’s uncertainty about whether or not there’s a negative effect.
If you actually take that systematic review seriously, the conclusion would be that philanthropists should fund more studies into investigating whether or not there’s an association given that our current evidence base is not enough to know.
Hm I don’t think that follows from the review- I would ideally like more studies looking at whether fluoride can affect IQ (esp at high concentrations), but I don’t think this should be the highest priority thing.
I want to highlight that the ‘low level evidence’ refers to fluoride at high concentrations. As I’ve outlined above, I think that fluoride interventions should only be used in areas with low fluoride levels. See the start of that review’s discussion, where it reads ‘This systematic review and meta-analysis gathered evidence showing that, following the WHO classification of low and high levels in the drinking water, exposure to low/adequate water F levels is not associated with any neurological damage, while exposure to high levels is. The level of evidence for this association, however, was considered very low.’
I could still see an argument to add in a risk factor to my CEA, but (bearing in mind that this is in a low fluoride area) I think this risk is sufficiently small that it is not worth including. For example, I haven’t included a factor for ‘not in pain = can go to school = higher IQ/ earnings’ which I’d argue has more support behind it. Nonetheless, I will keep an open mind and watch out for any new studies about this.
Instead of spending money on training local dentists, it might be more worthwhile to do disruptive innovation. There’s no reason to believe that the eyes of dentists are superior at diagnosis to a well-trained machine learning model.
Dental health care as currently practiced needs a lot of trust. I have to trust my local dentist with their assessment of whether my teeth need costly procedures. Having well-justified trust is easier in the first-world even when first world dentists still differ a lot in their treatment recommendations. In the third world it’s more likely that dentists will abuse trust by selling their customers unnecessary treatments.
If the diagnosis and treatment recommendations would be made by a branded computer program, it would get a lot easier to trust the model. Training costs would also go down if dentists wouldn’t need to be trained in diagnosis anymore.
Interventions could also plausibly be automated but that might be more expensive than just automating the diagnosis.
Hey Christian, thanks for your comment! I totally agree that ML has great potential for diagnosis (in dentistry but also within the field of medical care more broadly– e.g. I was reading about this grant from Gates the other day, it’s a diagnostic ultrasound for maternal conditions). Caveat that I’m not sure that the average person would trust an ML diagnosis over a ‘real person’ diagnosis (at least, not yet), and I think it would be a while until roll-out in an LMIC type context. Nonetheless, I think this is a promising area which I want to check out in more detail.
Within the field of oral health though, I think that investing in preventative treatment is currently more promising than diagnostics/ curative treatment: i.e. while I think that training up local dentists may be cost-effective, it seems a little borderline. However, preventative treatment (aka fluoride) seems to be extremely cost-effective, and we already know how to do it. This is mainly because salt and milk fluoridation appear to work and are insanely cheap—that’s where my $2 WELLBY/ $9 DALY figures are coming from.
Part of why it’s cheap is because you seem to put the non-monetary costs at zero.
Studies like https://ehp.niehs.nih.gov/doi/10.1289/ehp655 do suggest that there’s a risk of negative side effects from just putting fluoride into everything.
Teeth need fluoride but it’s likely harmful to the rest of the body. Application via toothpaste a lot more of the fluoride actually goes to the teeth has advantages over adding it to salt.
It might be that salt and milk fluoridation is still net-beneficial but that calculation is more complicated than just comparing the monetary cost.
As far as trust is concerned plenty of people also don’t trust water, salt or milk fluoridization and programs by Westerns in Africa might face political problems.
Hi Christian,
Thanks for this comment, I am happy someone drew attention to this!
I did check out various papers (such as the one you linked, and some other papers proposing a link between fluoride exposure and adhd). I didn’t find them convincing enough to add in a non-monetary cost to fluoride (bear in mind that I also left out various potential net-benefits, such as increased attendance at school and work).
To give a bit more detail (because I think this important), I found that the evidence linking fluoride to neurological outcomes was pretty low quality. I could only mainly find ecological studies, where i judge the risk of confounds to be very high- plus other issues that make me suspicious such as exclusion of outliers (without clear justification). That being said, I do think it’s important to be cognisant of these theories, even if the available evidence is poor— I quite like this review, which basically argues that low to adequate levels of fluoride don’t appear to be associated with any neurological conditions, but there is ‘low level’ evidence that they are at high concentrations.
To me, this gets at the point that fluoride should only be used in communities that have low levels of fluoride already— apart the speculation about neurological disorders, we know that fluoride can cause fluorosis (weakening of tooth enamel) in levels with high levels of fluoride. I have put the ‘costs’ of fluoride at zero, but have highlighted in the report that I think these kinds of interventions should only be used in area with low fluoride (for instance, not areas with high levels of groundwater fluoride). In areas with high fluoride, I agree with you that there might be ‘costs’ to be accounted for .
Totally agree with you on the point about trust! I quite like the salt/ milk interventions since they are more ‘opt in’ than water fluoridation (aka you could just buy the other salt/ milk), as well as being more practical in areas with lower levels of piped water access. I’d hope that this would mean less pushback.
The general way to account for side effects of clinical interventions is that low-level evidence is enough to take side effects seriously. Low-level evidence is enough for the FDA to require drug manufacturers to put a warning label on a drug.
The review you point to says “indicating the need for new epidemiological studies that could provide further evidences regarding this possible association”. That means there’s uncertainty about whether or not there’s a negative effect.
If you actually take that systematic review seriously, the conclusion would be that philanthropists should fund more studies into investigating whether or not there’s an association given that our current evidence base is not enough to know.
Hm I don’t think that follows from the review- I would ideally like more studies looking at whether fluoride can affect IQ (esp at high concentrations), but I don’t think this should be the highest priority thing.
I want to highlight that the ‘low level evidence’ refers to fluoride at high concentrations. As I’ve outlined above, I think that fluoride interventions should only be used in areas with low fluoride levels. See the start of that review’s discussion, where it reads ‘This systematic review and meta-analysis gathered evidence showing that, following the WHO classification of low and high levels in the drinking water, exposure to low/adequate water F levels is not associated with any neurological damage, while exposure to high levels is. The level of evidence for this association, however, was considered very low.’
I could still see an argument to add in a risk factor to my CEA, but (bearing in mind that this is in a low fluoride area) I think this risk is sufficiently small that it is not worth including. For example, I haven’t included a factor for ‘not in pain = can go to school = higher IQ/ earnings’ which I’d argue has more support behind it. Nonetheless, I will keep an open mind and watch out for any new studies about this.