Department of Health and Human Services
Food and Drug Administration
Office of Regulatory Affairs
FDA Geospatial Statistical Analyses (part deux):
Efficient Allocation of Real Property Across the
U.S.
The Idealist, Don Quixote, the Priest and Jiminy Cricket
We’re working on it….
Brian Holloway and Alexis Novoa
Introduction
To support the Food and Drug Administration’s (FDA) mission
many regulatory, investigative and science professionals are
assigned to a number of distinct geographic areas. The FDA real
property locations within these areas serve as administrative and
operational offices for our federal employees. Many of these
areas are experiencing shifting and expanding workloads based
on increased inspections of FDA regulated firms and products.
These “shifts and expansions” can benefit from geospatial
statistical analyses based on current and optimum real property
locations to better align workforce with locations to support the
FDA’s responsibility to protect the public’s health.
RPELM Methodology
In this study, the data set consisted of FDA visits
(inspections/investigations) over a 5 year cycle. The analyses
focused on a defined geographical area and compared the round
trip distance from the FDA location(s) within this area to the
“geospatial mean center” location within the defined
area/polygon.
The sample size of inspections totaled 205,000+ events. Each
event represented a distinct round trip allocation of FTE
resources to and from a site within the defined geographical
area. A standard descriptive statistical analysis was completed
comparing the FDA location(s) to the “geospatial mean center”
for each area and the resulting variances analyzed to determine
the “best location” for the area.
Part 1 of the study began in St Louis
RPELM Evolution
RPELM Evolution (cont.)
GIS Tools and FDA Inputs
1. User Inputs
1. Inspection/Investigation data – FDA db (.csv export)
2. OEI firm inventory – FDA db (.csv export)
3. FDA location(s) address(es) – current
4. Proposed new addresses based on “mean location”
5. Variance analysis
6. SharePoint 2010 – FDA internal
2. ESRI Tools
1. ArcGIS server (ver 10.2)
2. Portal for ArcGIS – FDA internal self-service GIS
3. ArcGIS Desktop (ver 10.2)
1. Network Analyst – closest facility tool, service area
2. Statistical Tool – mean location analysis
Summary of FDA Real Property
1. 5 Regions and 19 Districts
1. Encompassing 50 states and US territories
2. 235 Real Property Locations
1. 158 Resident Posts (Border Patrol/Point of Entry)
2. 20 District Offices
3. 13 Labs
4. Other: 40+ (parking garage, storage, guard shack)
3. Field FTE Count > 3,734
1. CSO/Investigators
2. SCSO
3. CO
4. Staff
RPELM “creating awareness”
1. Houston we have a problem (ESRI HHS Conf Atlanta Sep 2015)
How do we efficiently allocate/assign personnel and
renew/establish new locations across the U.S.?
Why Place Matters - Proposed Solution RPELM
model
1. Collect the 5 year inspection data and propose an
analysis based on “round trip” distance compared to the
“geospatial mean center” of an area
RPELM
Utah and Nebraska Locations Analysis
Mean Center and FDA location variance
Utah Mean Center
Total Distance to Firms
(round trip in miles)
n = 2006
Utah RP
Total Distance to Firms
(round trip in miles)
n = 2006
Nebraska Mean Center
Total Distance to Firms
(round trip in miles)
n = 1516
Nebraska RP
Total Distance to Firms
(round trip in miles)
n = 1516
190,061
180,860
298,809
297,787
Mean
4.59
Median
10.52
Nebraska Mean Center and
Nebraska RP
Variance
(round trip)
n = 1516
Mean
0.68
Median
13.30
Mode
39.85
Mode
-164.40
Standard Deviation
31.14
Standard Deviation
128.05
Minimum
-40.50
Minimum
-164.91
Maximum
43.26
Maximum
173.08
Utah Mean Center and Utah RP
Variance
(round trip)
n = 2006
Sum
Count
9,201.22
2006
Sum
Count
1,026.91
1516
Utah
Utah
Service Area Analysis
Drive Distance Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
Percent
1533
76.4
473
23.6
2006
100
Drive Time Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
1679
327
2006
Percent
83.7
16.3
100
Nebraska
Nebraska
Nebraska
Drive Distance Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
Percent
722
47.5
799
52.5
1521
100
Drive Time Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms Percent
830 54.6
691 45.4
1521 100
Montana Locations Analysis
Mean Center and FDA location variance
Montana Mean Center
Total Distance to Firms
(round trip in miles)
n = 1210
Montana RP
Total Distance to Firms
(round trip in miles)
n = 1210
Montana BS
Total Distance to Firms
(round trip in miles)
n = 1210
447,193
404,665
556,015
Montana Mean Center and Montana
RP
Variance
(round trip)
n = 1210
Mean
35.16
Montana Mean Center and Montana
BS
Variance
(round trip)
n = 1210
Mean
-89.94
Median
53.60
Median
-71.60
Mode
289.68
Mode
-414.91
Standard Deviation
202.99
Standard Deviation
218.37
Minimum
-303.38
Minimum
-422.09
Maximum
307.18
Maximum
411.25
Sum
42,511
Sum
Count
1210
Count
-108,822
1210
Montana
Montana
Montana
Drive Distance Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
113
1097
1210
Percent
9.3
90.7
100
Drive Time Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
232
978
1210
Percent
19.2
80.8
100
Idaho Locations Analysis
Mean Center and FDA location variance
Idaho Mean Center
Total Distance to Firms
(round trip in miles)
n = 1412
Idaho RP
Total Distance to Firms
(round trip in miles)
n = 1412
Idaho BS
Total Distance to Firms
(round trip in miles)
n = 1412
617,170
425,507
1,356,419
Idaho Mean Center and Idaho RP
Variance
(round trip)
n = 1412
Idaho Mean Center and Idaho BS
Variance
(round trip)
n = 1412
Mean
97.14
Mean
-515.88
Median
76.82
Median
-646.95
Mode
76.82
Mode
-837.48
Standard Deviation
142.87
Standard Deviation
451.78
Minimum
-277.19
Minimum
-837.48
Maximum
280.51
Maximum
807.33
Sum
Count
137,168.18
1412
Sum
Count
-728,426.75
1412
Idaho
Idaho
Idaho
Idaho
Drive Distance Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
573
839
1412
Percent
40.6
59.4
100
Drive Time Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
643
769
1412
Percent
45.5
54.5
100
Kansas Locations Analysis
Mean Center and FDA location variance
Kansas Mean Center
Total Distance to Firms
(round trip in miles)
n = 1510
Kansas OCI
Total Distance to Firms
(round trip in miles)
n = 1510
KAN DO
Total Distance to Firms
(round trip in miles)
n = 1510
Wichita RP
Total Distance to Firms
(round trip in miles)
n = 1510
625,354
479,654
467,501
528,024
Kansas Mean Center and Kansas OCI
Variance
(round trip)
n = 1510
Kansas Mean Center and KAN DO
Variance
(round trip)
n = 1510
Kansas Mean Center and Wichita RP
Variance
(round trip)
n = 1510
Mean
72.31
Mean
78.34
Mean
48.30
Median
189.98
Median
196.53
Median
77.22
Mode
-269.94
Mode
-260.04
Mode
87.53
Standard Deviation
321.54
Standard Deviation
318.26
Standard Deviation
99.43
Minimum
-441.88
Minimum
-431.98
Minimum
-196.16
Maximum
449.91
Maximum
440.01
Maximum
196.23
Sum
Count
145,700.10
1510
Sum
Count
157,853.07
1510
Sum
Count
97,330.19
1510
Kansas
Kansas
Kansas
Kansas
Drive Distance Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
Percent
1146
56.8
870 43.2
2016
100
Drive Time Coverage
Inside Coverage Area
Outside Coverage Area
Total
Firms
1494
522
2016
Percent
74
26
100
Proposed Solution RPELM model
1. Collect the 5 year inspection data and propose an
analysis based on “round trip” distance compared to the
“geospatial mean center” of an area
2. Rerun analysis based on “drive time” coverage from
existing FDA locations (50 and 100 mile coverage)
3. Rerun analysis based on county boundaries
disregarding state and FDA district boundaries (that’s a lot of
work…may need a python dev team, javascript programmer and GIS consult)
…..stay tuned
FDA SharePoint Map – (html popup)
CONTACTS
Department of Health and Human Services
Food and Drug Administration
Office of Regulatory Affairs
Brian Holloway
Alexis Novoa
Darren Blumberg
Joseph Mastrianna
[email protected]
[email protected]
[email protected]
[email protected]