Metrics - 24 hour mean or annual mean

Hi everybody!!

My question is:

I have to use the same metrics at epidemiologic study and air quality data, I mean, if the epidemiologic evidence analyses 24 hour air quality variations, so at benmap map i have to use 24 hours means of air quality, is it correct? If so, then my question is how to considering 24h means variations if BenMap calculates a mean of all values to determine air quality decrement?

If I not clear, my question is:
When using beta coeficients values that derives from epidemiologic studies regarding daily variations of polluants, when BenMAP, using quality data from monitors, calculates a mean from these values it doesn’t considers daily variations that are greater than annual mean variations? In these case doesn’t matter if I input daily values (24 hours mean) or annual mean values? How can I compute daily variations at my analysis?

Hi Alex,

Thanks for your question and your interest in BenMAP-CE.

Your question is clear. You are correct. Using an epidemiological study reporting risk coefficients for a 24-hour change in air pollution will generally require users to import 24-hour average concentrations to BenMAP-CE. However, you do not need to import 365 individual daily values, but rather than import an annual average of daily values. If you import annual mean values, the program will use this annual mean as a surrogate for the mean (i.e., it will use the annual mean value for each of 365 days of the year).

I hope this helps. Good luck with your project!

-Neal

Ok, I understand your point. My doubt point was because in manual of BenMAP, at 4-14 page, was write like that "…“the air quality change must be expressed in a metric that matches the metric used by health impact function”, In true this sentence concerns about how the metric is defined in program input, but not as the epidemiologic study was developed, is it correct?

That said it doens’t mather if I used daily values (24 hour mean) or annual mean like input at BenMAP, the software always works at annual mean base, correct?

Considering that 24 hours mean variations are greather than annual mean variations, there is a way to utilize that variations in BenMAP analyses, rather than annual mean variations? This is a limitation of program?

Hi Alex,

That statement is mostly correct. The air quality metric in the epidemiological study supplying the risk coefficient should match the air quality metric in the health impact function calculated by BenMAP-CE.

If you supply the software with a daily mean for each day of the year, it will calculate a health impact estimate for each day, using each of 365 values. If you supply it with a single annual mean, it will use this annual mean for each of 365 days.

For log-linear functions, the results will be very similar using either a daily mean or annual mean.

Thanks again for your support. What you said appears clear, but in pratice… Let me explain: In my case analyses I am using monitor data values from air quality stations. Once we don’t have stations for all points of area, BenMAP uses a interpolation method to assign air concentrations for all points, from where air increment quality will be estimated (baseline - control). Considering 365 values of daily mean that I input at program for all stations, Does BenMAP simply calculates a average from values (similar to annual mean) for each station and interpolate for all area?? In this case we don’t have analyses for each day of year, but only for a mean annual average?
Do you have any other support material for consult that explain better this issue?

Regards

Hi Alex and Dr. Fann,

I met similar question. Though Dr. Fann has clarified some questions, I still have some concerns:

  1. If my modeled PM2.5 is 24 hour average basis (365 daily mean), but the incidence rates we got is annual average (1 for each location), then how did the BenMAP process it. For example, using Bell’s (2012) study, we can estimate the hospital admissions for cardiovascular disease. In theory, if I understand correctly, we should use daily changes in PM2.5 and corresponding daily incidence rate to calculate cardiovascular hospital admissions for each day. However, the preloaded incidence rate (Other Incidence 2014) data are annual-basis. In this case, does the BenMAP use annual mean incidence as a surrogate for daily incidence rate and daily PM2.5 to compute the health impact. Or, BenMAP calculates annual mean PM2.5 first and then use annual mean PM2.5 and annual-basis incidence rate to quantify the short-term impacts?

  2. Like what Alex asked, if there are missing values, how did the BenMAP process these data. Simply average them and treat them as annual average?

Thank you,

Best,
Xiangyu

Hi BenMAP users,

Anyone can shed light on this?

Thank you :slight_smile:

Best,
Xiangyu

Hi Xiangyu,

I’m sorry for the delay in responding. I will answer your questions in order:

  1. In short, the incidence rates should be expressed at the same temporal scale as the epidemiological study. In your example, the baseline rate of hospital admissions should be expressed as per person per day. All of the incidence rates in BenMAP-CE, except for mortality rates, are calculated this way. The mortality rates are expressed as per person per year
  2. If you supply BenMAP-CE with 365 daily values, it will calculate impacts for each day of the year using each of the 365 values. If you supply the program with a single annual mean, then it will use this single annual value as a surrogate for the daily change.

I hope this helps!

-Neal

Hi Neal,

So, for each cell of grid analyses BenMAP, considers all variations for all days (24 hour means values), then inform the attributabble number of cases at day bases (cases per day). The point estimate of results is the mean of normal distribution that considers all the 365 increments values?
If I want a annual base of attributabble number of cases for all grids cells, Can I just sum then and multiply to 365?

Best

Hi Dr. Fann,

Thanks for your reply! I have one more question regarding to this issue. :slight_smile:

In my case, I supply 365 daily PM2.5 values from CMAQ model (4km) and BenMAP provides one baseline rate of hospital admissions for each county. You mentioned “If you supply BenMAP-CE with 365 daily values, it will calculate impact for each day of the year using each of the 365 values”. If I understand correctly, it should provide 365 daily change outputs for each grid (e.g, 4km). However, I only got one value for each grid. How to interpret this output? Is it averaged hospital admissions daily change? Why not 365 outputs but only one single value for each user-defined grid? I am a little bit confused.

Thank you!

Best,
Xiangyu

Hi Xiangyu,

Trying to help you about your issue, as it was my question too, the single value informed by BenMAP is a point estimate, that is, is the mean value got from 365 daily values attributed health effect change calculated for each cell grid. I think that distribution around the mean (point estimate) informed as percentiles computes all daily variations, but I still waiting for confirmation from EPA’s experts.

Best,

Alex and Xiangyu – If your health impact function (HIF) uses an annual statistic, such as the annual mean concentration, the resulting point estimate provided by BenMAP-CE is based on the health impacts associated with the change in that summary measure (e.g., the change in the annual mean). If you have entered 365 daily air quality values into BenMAP-CE, the program will calculate the annual statistic from all of those values before calculating the HIF. If the health impact function you are using relies on daily incidence rates (e.g. those for ER visits and hospital admissions in the US setup); then BenMAP-CE will calculate the HIF results for each day based on the change in pollutant concentration each day in each grid cell; BenMAP will then sum these daily estimates into one annual estimate for each grid cell and then sum those annual results across all grid cells. The annual sum of the health impacts is what BenMAP reports as the point estimate. It does not report the individual daily changes.
If you are unclear about whether your HIF uses a daily or annual pollutant concentration, you can go to Modify Datasets, click Manage under Health Impact Function, then Edit the Dataset and Function of interest and look at the values under “Metric:” and “Annual Statistic:”. (See screenshot.)


If Annual Statistic says “None”, then the function uses daily values.
Xiangyu – You may consider consulting Appendix B of the BenMAP-CE user manual for further information on the averaging process.
Also note that the distribution around the point estimate reflects uncertainty in the concentration-response function parameters estimated in the epidemiological study and uncertainties in the valuation estimates if you are using any. The distribution does not reflect the variability in the pollutant concentrations.

Ok! Henry, thanks for your support. But looking to my case, I have entered 365 daily values into BenMAP, and set my HIF to using D24hour mean as metric. If I run software like this, using daily incidence rates, the point estimate informed by BenMAP is an annual bases (the sum of all days estimates at each cell grid), right?
However if I don’t set annual statistic as “mean” my incidence results appear null, that is the BenMap doesn’t calculates attributable incidence. What’s can be wrong with this analyses??

Best,

Alex,

I think BenMAP does not recognize your air quality concentrations as daily values. In your air quality monitor data input files, what values are you using for the columns Metric, Seasonal Metric, and Statistic?

Hi Henry,

metric = D24HourMean
Seasonal Metric = “blank”
Statistic = “blank”

As show below for one air quality station of my study:

Monitor Name,Monitor Description,Latitude,Longitude,Metric,Seasonal Metric,Statistic,Values
1,"‘STATION1’","-20.35272196148703557","-40.56344540683321071",D24HourMean,“5.53,10.33,9.38,6.00,5.34,5.22,3.39,3.93,5.13,6.71,6.61,7.96,6.88,5.31,6.28,4.57,4.71,6.17,6.08,5.07,7.43,5.95,6.32,5.22,4.49,7.36,7.02,5.70,9.40,17.07,8.06,6.44,4.89,6.42,6.96,7.07,7.62,5.80,4.66,6.52,4.36,5.34,5.58,.,.,.,.,5.24,8.43,6.04,4.34,6.22,4.26,3.77,4.34,4.91,6.00,6.55,11.01,4.38,2.84,3.27,6.64,4.83,5.08,5.49,4.44,.,.,.,.,5.68,6.47,6.19,8.33,7.83,6.74,5.40,5.60,5.34,6.73,2.57,4.98,3.00,4.14,10.34,14.76,7.66,5.61,8.07,8.18,6.70,.,5.66,4.57,7.99,5.26,4.04,4.67,3.27,3.21,3.42,6.56,5.92,8.82,6.55,10.92,8.72,8.76,9.66,11.95,.,3.73,4.29,6.02,3.72,3.31,6.67,5.22,2.89,2.43,3.97,4.61,7.78,6.63,2.98,2.28,3.68,3.72,4.74,6.22,4.85,5.60,2.14,3.37,1.83,2.13,4.84,2.60,2.94,4.32,3.91,4.65,5.73,6.01,11.26,11.41,8.76,6.15,6.54,8.53,9.40,3.85,6.62,4.20,6.08,5.76,4.33,5.07,5.83,9.24,9.71,9.62,9.37,5.80,11.17,2.15,5.33,12.23,6.85,.,3.00,3.46,8.42,6.66,.,2.30,2.25,1.83,3.79,6.27,7.56,8.72,7.63,6.46,4.76,2.92,4.16,6.16,13.42,12.34,8.15,5.35,5.14,5.21,6.03,4.87,5.72,5.72,4.04,4.63,6.99,5.88,3.52,4.16,6.04,6.18,7.10,7.71,9.30,7.65,6.22,5.99,3.47,5.95,8.60,10.54,6.53,6.28,4.25,2.95,2.90,2.86,3.12,3.62,4.08,3.41,3.87,5.43,4.68,7.24,8.50,8.07,6.11,7.03,15.41,13.09,.,.,11.22,2.04,2.11,2.37,2.85,3.65,3.28,3.30,3.27,2.10,2.83,3.25,5.33,6.55,8.37,4.61,3.00,2.54,3.38,5.34,5.30,4.74,5.04,5.16,7.21,5.49,6.73,11.73,11.98,12.44,8.62,7.44,9.71,10.64,11.48,8.89,8.02,7.90,2.48,3.58,2.44,3.36,4.43,5.28,7.71,6.65,9.76,10.82,14.32,12.43,13.84,5.92,2.80,5.16,9.89,.,19.22,10.15,4.95,6.14,9.24,10.90,6.53,4.17,7.50,9.30,8.66,11.88,14.20,15.02,18.73,16.33,11.61,11.34,8.41,7.17,13.39,8.39,16.20,17.57,4.54,6.63,10.38,13.38,14.76,9.68,3.27,5.78,13.03,8.23,4.00,3.57,5.17,7.64,11.15,4.74,6.52,9.07,4.00,10.26,11.83,7.37,3.32,8.18,11.87,20.70,8.76,4.78,5.40,4.15,6.02,6.34,7.89,4.45,4.48,4.84,3.68,3.70,5.15,4.81,5.57,6.71,11.25,10.39,6.43,7.89”

What do you think?

about this setence

_But looking to my case, I have entered 365 daily values into BenMAP, and set my HIF to using D24hour mean as metric. If I run software like this, using daily incidence rates, the point estimate informed by BenMAP is an annual bases (the sum of all days estimates at each cell grid), right?

What do you think?

best

If all of your monitor entries look like the example you posted, then the problem is not with your air quality input file.

If I understand you correctly, if you specify your HIF to use daily values (Annual Statistic = “None”) you get no results (all zeros), but if you specify HIF to use the Annual Statistic = Mean, BenMAP generates positive results. Is that right?

What concentrations are you using for your control scenario? Are you doing a monitor rollback or specifying your own values?

If I understand you correctly, if you specify your HIF to use daily values (Annual Statistic = “None”) you get no results (all zeros), but if you specify HIF to use the Annual Statistic = Mean, BenMAP generates positive results. Is that right?

THAT’S RIGHT!

Control and baseline scenarios are from monitor data (from air quality stations located in my area of study). I’m comparing two years of data (year 1 = baseline and year 2 = control)

Alex- sorry for the difficulties you are encountering with BenMAP.

To better help you with this issue, would you be willing to share your input files? If so, you may either post them here (publicly) or email them to me directly at wraich@indecon.com.

It sounds like this may be a bug that our programmers can attempt to address.

Thanks,

Will

Hi Henry and Alex,

Thanks very much for your help! It’s very clear now! :slight_smile:

Best,
Xiangyu