Tuesday, January 31, 2017
Saturday, January 28, 2017
Ruth Fairbank, M.D., The First Female Neuropsychiatric Epidemiologist?
Blog readers know I'm trying to remedy general neglect of biographies of early and eminent women who were leaders in our field. I've had a longtime hypothesis that Ruth Fairbank was the first, based on some of my interviews with Paul Lemkau in the late 1970s during lunchtime history sessions in the wonderful 9th Floor Cafeteria of the JHPH Wolfe Street building. Paul described her as a compassionate intelligent psychiatrist whose leadership was crucial to the success of the Eastern Health District surveys that Adolph Meyer wanted to be carried out in a collaboration of the Phipps Clinic and the School of Hygiene and Public Health, each unit of the university positioned no more than a very short walk across Wolfe Street.
I have found a number of sources, but none better than Elizabeth Fee's paragraphs in her history of the SHPH, reproduced here in an effort to get you to buy and read her book.
Pay attention to the spot-mapping approach.
Sunday, January 15, 2017
Female-male difference in fatality associated with vehicle-pedestrian collisions
Crash fatality decomposition
This decomposition into stages serves as a great example of epidemiology' fourth rubric (Mechanisms).
Friday, January 13, 2017
National Academies light up cannabis, again...
https://www.nap.edu/read/24625/chapter/1
Admirable endeavor of considerable value, but some lapses in the form (on occasion) of pedestrian unthoughtful thinking, or perhaps shallow would be a better description of the thinking. Standard "more research is needed" recommendation as might be expected in a National of Academies report, but many unrealistic and uncritical assertions about what is needed. The profile of general recommendations might be a consequence of the "general epidemiology" composition of the team, and little field experience in cannabis research, despite the stellar credentials of all panelists, and the deep depth experience of some of the cannabis researchers (e.g., N. Kaminski).
Something along the lines of this statement recurs:
"No or little systematic review of the published evidence was found."
A requirement for a systematic review is a sufficient number of controlled studies to create a chain of inference, and what exists at present is a set of links of highly uneven quality, and generally too insubstantial to support systematic review.
Why?
During the early 1980s, Mary Monk and I learned of lukewarm NIDA study section enthusiasm for our proposed NIDA epidemiological research on cannabis problems, which had focus on creation of primary care networks for systematic ascertainment of clinically significant cases of what now is called "cannabis use disorder." (See how we adapted this community practice study design for the first "non-specialty clinic" case-control study of the dementia syndromes, in our Sydney, Australia research, Henderson et al., 1992.) That is, there was little wrong about study design or logic. Rather, the study was belittled for its focus on cannabis problems. Later, a member of that study section took me aside and counseled me to start studying "serious drug problems," by which he meant what he was studying -- namely, heroin. Lesson: what looks like the jugular to you can look like a capillary to others.
But there is a more fundamental problem in that the community-based practice research concept we developed was one that allows replication because the design is relatively inexpensive and did not require the more expensive apparatus of drawing a probability sample from a defined study population and then repeated longitudinal measurements, which seems to have been the gold standard for evidence applied here.
Admirable endeavor of considerable value, but some lapses in the form (on occasion) of pedestrian unthoughtful thinking, or perhaps shallow would be a better description of the thinking. Standard "more research is needed" recommendation as might be expected in a National of Academies report, but many unrealistic and uncritical assertions about what is needed. The profile of general recommendations might be a consequence of the "general epidemiology" composition of the team, and little field experience in cannabis research, despite the stellar credentials of all panelists, and the deep depth experience of some of the cannabis researchers (e.g., N. Kaminski).
Something along the lines of this statement recurs:
"No or little systematic review of the published evidence was found."
A requirement for a systematic review is a sufficient number of controlled studies to create a chain of inference, and what exists at present is a set of links of highly uneven quality, and generally too insubstantial to support systematic review.
Why?
During the early 1980s, Mary Monk and I learned of lukewarm NIDA study section enthusiasm for our proposed NIDA epidemiological research on cannabis problems, which had focus on creation of primary care networks for systematic ascertainment of clinically significant cases of what now is called "cannabis use disorder." (See how we adapted this community practice study design for the first "non-specialty clinic" case-control study of the dementia syndromes, in our Sydney, Australia research, Henderson et al., 1992.) That is, there was little wrong about study design or logic. Rather, the study was belittled for its focus on cannabis problems. Later, a member of that study section took me aside and counseled me to start studying "serious drug problems," by which he meant what he was studying -- namely, heroin. Lesson: what looks like the jugular to you can look like a capillary to others.
But there is a more fundamental problem in that the community-based practice research concept we developed was one that allows replication because the design is relatively inexpensive and did not require the more expensive apparatus of drawing a probability sample from a defined study population and then repeated longitudinal measurements, which seems to have been the gold standard for evidence applied here.
All well and good, but the nation's current investments in longitudinal PATH and ABCD studies are not going to yield replications. Designed as massive longitudinal studies, the designs do not set up the multiple replications required to produce multiple links in any chain of inference. Instead, a single study estimate will be produced and that again means not enough evidence for systematic review. Ten years from now, that part of the critique will be the same, when a new NAS panel is convened.
If I have time, I might write some more on this topic of NIDA's relative neglect of relatively inexpensive multi-site studies, each of which "has its own bottom" and can produce a useful estimate in a series of systematic replications, versus its adoption of the massive study approach under the cooperative agreement model, which yields one estimate.
Perhaps someone can comment on a topic I did not have a chance to check. Did they interview or hear presentations from ABCD leaders on what ABCD is doing about measurement of dose, route of administration, between-dosing intervals, etc., of the type that must be faced when studying cannabis in a large sample longitudinal real world context? The panel recommendations about the field's neglect of these variables belies (a) a lack of familiarity with the specific context of cannabis research when DEA has driven the behavior underground, and pretty much did exactly the opposite of what the National Commission on Marihuana and Drug Abuse recommended in the early 1970s? (Side-note: check Senator Sessions' testimony and actions on the federal law enforcement front; back to those "culture wars" later), and (2) practical problems faced in a regulatory environment that thwarts standardization of bioavailability and bioequivalence measurements. Perhaps the best that now might be done in cannabis research is reflected in what the PATH study is able to do for tobacco and nicotine delivery products, which have been more tightly regulated, and have had the advantage of more than a half-century of well-funded NCI and other NIH research. Someone might have contrasted the number of total research dollars allocated across NIH institutes for tobacco and nicotine with the comparable dollar amount for cannabis, and reflect on what should be realistically expected. For example, in a thought experiment, go back to the year of NIH dollar-adjusted "tobacco/nicotine health hazards" epidemiology expenditures when the NIH reached the dollar amount that it has been spent specifically on "cannabis health hazards" epidemiology research (NIH-wide), to date. Then, check the quality and nature of epidemiological evidence on tobacco/nicotine health hazards at that time, in that year, relative to the quality and nature of epidemiological research on cannabis health hazards at present. E.g., were there systematic reviews of tobacco health effects in that year? I suspect not.
But at the end of the day, the NAS panel has done a service by pointing out that more cannabis research is needed, and they were not silent on the dampening effect of federal policy when it comes to learning about the health hazards of cannabis (which I study) or the potential health benefits (which I encourage others to study).
They had to say something, and in general, what they said was not bad; it was just unrealistic.
[Postscript: The cannabis research dollar total should not include the Monitoring the Future study, which represents NIDA's most substantial investment in study of adolescent and young adult drug use, but the study is psychosocial research in its orientation, and never has produced authoritative evidence on cannabis hazards, due to its psychosocial research orientation. It has and still can produce useful estimates on other things (e.g., how many kids using each year, provided they stay in school; cannabis risk perceptions; school dropout), but it's a pretty meager source of evidence on health hazards, either cross-sectionally or longitudinally, with its longitudinal evidence severely constrained by massive sample attrition and mailed questionnaire measurements. we cannot expect too much in the way of definitive epidemiological evidence on cannabis or other drug health hazards when there is so much missing data in the longitudinal trace, and no serious attempt to assess validity of the mailed questionnaire responses about health status).]
Apologies for typos and possibly an errant URL. Moving fast today.
Tuesday, January 10, 2017
Articulate bioethicist experience with opioid withdrawal
Bioethicist withdrawal story
Health Affairs companion article.
Health Affairs companion article.
The Cartesian mind-body dualism is surprising.
Some lack of insight about how the constructs of "psychological" dependence dovetail with "biological dependence," and the distinction has been abandoned in research for lack of utility.
But some value must reside in the dualism since it lives on.
People vary considerably in the experience of opioid withdrawal.
Many have brushed it off as a relatively trivial experience (but for gut problems).
His experience was more extreme, at the other end of the distribution.
Sunday, January 8, 2017
Vermont "weed dating"
Considering what Jacob found out about Vermont in his cannabis incidence mapping project, this is not what I expected to read about in this article:
In 2025, who will be eating whose lunch?
Data Science morphs out of epidemiology and biostatistics.
(I've been watching the trend for the past 5-8 years. This note is from 2013. Not hard to see where it's going. Try to integrate thinking on this trend with recently circulated article on boundaries of epidemiology in the 20th century.)
Our field continues to focus on narrow-band scientific throughput while data sciences expand bandwidth. Think of that EHR-based case-control study you designed for class as narrow band. Now, re-conceptualize with the frame that a pharma market analyst might bring into play. "Can I use all the information of the health care sector, not just the Dx code in EHR, but also pharmacy records, lab tests, imaging orders, etc., thinking of them as pipes (bands) of information I can integrate to identify unrecognized (and from sales volume standpoint, the not yet treated) cases lurking in the "control" stratum of that study base?"
That is, the problem is not set up to be satisfied with what the EHR Dx code says, by itself. The problem is set up to detect a latent class of potential consumers from which next year's Dx codes either will show zeroes (the incipient case remains unrecognized for yet another year) or ones (bingo! recognition of the Dx indication, without which the sales volume remains flat).
Should you be making plans for a summer internship or sabbatical at Merck? Or Alkhermes?
(I've been watching the trend for the past 5-8 years. This note is from 2013. Not hard to see where it's going. Try to integrate thinking on this trend with recently circulated article on boundaries of epidemiology in the 20th century.)
Our field continues to focus on narrow-band scientific throughput while data sciences expand bandwidth. Think of that EHR-based case-control study you designed for class as narrow band. Now, re-conceptualize with the frame that a pharma market analyst might bring into play. "Can I use all the information of the health care sector, not just the Dx code in EHR, but also pharmacy records, lab tests, imaging orders, etc., thinking of them as pipes (bands) of information I can integrate to identify unrecognized (and from sales volume standpoint, the not yet treated) cases lurking in the "control" stratum of that study base?"
That is, the problem is not set up to be satisfied with what the EHR Dx code says, by itself. The problem is set up to detect a latent class of potential consumers from which next year's Dx codes either will show zeroes (the incipient case remains unrecognized for yet another year) or ones (bingo! recognition of the Dx indication, without which the sales volume remains flat).
Should you be making plans for a summer internship or sabbatical at Merck? Or Alkhermes?
Example of integrated comparative effectiveness research with sensitivity analyses for possible selection effects; deserves study as an early example of CER in post-PCORI era.
Saturday, January 7, 2017
Heroin: Szasz, Densen-Gerber, Dupont, Jaffe, and Nixon's war on drugs
Don't believe everything you read.
See early photo of Dr. Jerome Jaffe, our first "drug czar," who appears on the right side of the photo
1971 article
See Densen-Gerber anecdote of person-to-person spread in Utah within the article, as well as Szasz critique.
See early photo of Dr. Jerome Jaffe, our first "drug czar," who appears on the right side of the photo
1971 article
Trying to Treat the 'White Plague'
Science News 1971, get through JSTOR
See Densen-Gerber anecdote of person-to-person spread in Utah within the article, as well as Szasz critique.
Friday, January 6, 2017
Marijuana Anonymous and a question about alcohol.
Link to interesting article by a smart journalist who interviewed me by phone over the holiday (and mainly got it right). More right than the RollingStone writer who "stepped on" my suggestion that a parent contemplating purchase of a drug test kit should instead use the cost of the test kit to take the kid out to dinner and just listen, lisren, listen
Is there a Marijuana Anonymous Chapter near you?
The article cites Catalina's NESARC study that I told Shelby to check out, and it now prompts me to ask:
Why did the NIAAA in-house study of transition probabilities overshoot our NIMH study estimate that about one in 7-8 drinkers would be affected by alcohol dependence?
The NIAAA estimate places EOH up there with cocaine and bumps our 1in 3 tobacco estimate to 2 in 3?
Who yah gonna trust?
Hoping for comment from Catalina or one of the NESARC investigators.
I suspect a "gate" in the diagnostic workup or in the analysis step such that NESARC used a different denominator. Possibly an artifact induced by tracking the survival function out to a very thin sample space on the right hand side of the curve.
Comment if you can narrow down the possibilities.
Is there a Marijuana Anonymous Chapter near you?
The article cites Catalina's NESARC study that I told Shelby to check out, and it now prompts me to ask:
Why did the NIAAA in-house study of transition probabilities overshoot our NIMH study estimate that about one in 7-8 drinkers would be affected by alcohol dependence?
The NIAAA estimate places EOH up there with cocaine and bumps our 1in 3 tobacco estimate to 2 in 3?
Who yah gonna trust?
Hoping for comment from Catalina or one of the NESARC investigators.
I suspect a "gate" in the diagnostic workup or in the analysis step such that NESARC used a different denominator. Possibly an artifact induced by tracking the survival function out to a very thin sample space on the right hand side of the curve.
Comment if you can narrow down the possibilities.
Wednesday, January 4, 2017
Monday, January 2, 2017
Apropos prior post on unusual syndromes
Cyclic vomiting syndrome and cannabis
Cries out for case-crossover, unless it's akin to a slow virus.
I'd be suspicious about uncontrolled confounding or contaminated product.
Cries out for case-crossover, unless it's akin to a slow virus.
I'd be suspicious about uncontrolled confounding or contaminated product.
Sunday, January 1, 2017
Intersections of Epidemiology with Demography
Trying to learn some more about counties of the Upper Midwest, I stumbled across this interesting US Census blog and an entry on age pyramids. Great overview.
Blog on age pyramids
Child poverty trends: dynamic mapping from 2007-2015
Other county level data
Blog on age pyramids
Child poverty trends: dynamic mapping from 2007-2015
Other county level data
Subscribe to:
Posts (Atom)