Opioid Vulnerability in Massachusetts Spatial epidemiological analyses to inform Massachusetts public health officials Authors: Liz Marsh, Sumeeta Srinivasan, Erin Jaque, Len Young, Thomas Stopka (P.I.)

Vulnerability maps were calculated by rasterizing vector data. Fatalities were converted from counts to rates by dividing by Zip Code Tabulation Area (ZCTA) population counts from the American Community Survey 2016 estimates of population aged 10 and older. Using Inverse Distance Weighting (IDW), rates were converted to raster. IDW was also used for all PIP variables. IDW outputs were reclassified into 10 quintiles. High rates of fatalities and high rates of PIP were assigned higher scores, lower rates were assigned lower scores. Resource variables were converted using Euclidean Distance and reclassified into 10 quin-tiles. Low values were assigned for proximity to resources and high values were father away. For the overall vulnerability map, all reclassified variables were combined using Weighted Overlay. All variables were given equal weight. Zonal Statistics tool, which converts raster values to polygons were used to calculate vulnerability scores at the town level for all three score maps.

Overall Vulnerability: All Variables

#1 Drug overdose has become the leading cause of accidental death in the US1.

3x Opioid prescription rates have increased nearly threefold in the last 15 years

in the US—alongside an increase in opioid related overdoses and deaths2.

38% Of the estimated 52,404 lethal drug overdoses in 2015, 20,101 were related

to the use of prescription pain relievers2.

Overall vulnerability: Score by Town

Low Resources, High Fatality Rates Resource Vulnerability by Municipality

Background

Local Outlier Analysis of Opioid Overdose Fatalities

Methods: Point data for all opioid overdose fatalities from 2015-

2017 (not adjusted for population) were first run through a time-

space cube. The cube was used to run local outlier analysis with

a 5 kilometer fishnet grid and a 3 month time stamp.

Results: Local outlier analysis run on point data showed persis-

tent hotspots in and around Boston. It also showed many statis-

tically significant high-low spots scattered around the state.

Emerging Hotspot Analysis of Opioid Overdose Fatalities

Methods: Emerging hotspot analysis (Getis-Ord Gi*) was conducted

using the same space-time cube created for local outlier analysis, with

a 5 km fishnet and a 6 month time stamp. Results: Statistically signifi-

cant hotspots were found in the Boston area, mostly as intensifying hot

spots. South of Plymouth some significant consecutive and sporadic

cold spots were found. This preliminary analysis shows the scope of the

problem: deaths are increasing in the greater Boston area.

Which communities are experiencing the highest rates of opioid overdose fatalities, have low resources to address Opioid Use Disorder and have high rates of Potentially Inappropriate Prescribing?

Decision makers have responded to the opioid crisis with new policies and programs that provide resources in some places

where there are high rates of opioid overdose, but a timely analysis of communities in Massachusetts that have a rising

incidence of opioid fatalities but still lack resources will be useful to the Massachusetts Department of Public Health and

municipalities across the commonwealth, to make decisions about where to concentrate resources in Massachusetts.

*Data Sources: Massachusetts Department of Public Health, Prescription Monitoring Program; Bureau of Vital Statistics; Substance Abuse and Mental Health Services Administration | Funded by the Massachusetts Department of Public Health

Projected Coordinate System: NAD_1983_StatePlane_Massachusetts_Mainland_FIPS_2001| Projection: Lambert Conformal Conic | UEP 102 Advanced GIS, Spring 2019 | 7 May 2019 | References: 1) Schuchat A, Houry D, Guy GP. New data on opioid use and prescribing in the United

States. JAMA. 2017;318:425–426. 2) Jones CM, Paulozzi LJ, Mack KA. Sources of prescription opioid pain relievers by frequency of past-year nonmedical use United States, 2008-2011. JAMA Intern Med. 2014;174:802–3. 3) Rose, A. J., Bernson, D., Chui, K. K. H., Land, T., Walley, A. Y., LaRo-

chelle, M. R., & Stopka, T. J. (2018). Potentially inappropriate opioid prescribing, overdose, and mortality in Massachusetts, 2011–2015. Journal of general internal medicine, 33(9), 1512-1519. 4) Stopka, T. J., Amaravadi, H., Kaplan, A. R., Hoh, R., Bernson, D., Chui, K. K., ... & Rose, A. J.

(2019). Opioid overdose deaths and potentially inappropriate opioid prescribing practices (PIP): a spatial epidemiological study. International Journal of Drug Policy, 68, 37-45. *Service Layer Credits: Esri, HERE, Garmin © OpenStreetMap contributors and the GIS user community. |

Symbols: https://svgsilh.com/image/2028286.html , https://pixabay.com/vectors/drug-icon-pill-icon-medicine-icon-2316244/ , https://commons.wikimedia.org/wiki/File:Hospital_font_awesome.svg

Fatality Variable3,4,* Opioid Overdose Fatalities: Address of death for all opioid overdose fatalities in Massachusetts from 2015-2017.

PIP Definition3,4,* Potentially Inappropriate Prescribing – a pattern of opioid prescribing by medical professionals that may be a risk factor for fatal overdose. Definitions for each measure of PIP are described below.

PIP Variables 3,4,* Pip1: High-dose opioids. Daily dose equivalent for each patient in each month in MME. Patients received high-dose opioids if they

had an MME > 100mg/day in three separate months. Pip2: Overlapping opioid and benzodiazepine prescriptions. Defined as having received overlapping pre-scriptions if there were at least three months when the patient had received opi-oid and benzodiazepine prescriptions that overlapped by at least one day. Pip3: multiple opioid prescribers. Number of patients with 4 or more opioid prescrib-ers in any quarter. Pip4: multiple opioid pharmacies. Number of patients with 4 or more pharmacies in any quarter. Pip5: cash purchases of prescription opioids. If patient paid cash on three or more separate occasions during the time period.

Resource Variables for Massachusetts*: Naloxone Distribution: Address-es of all Overdose Education and Naloxone Distribution Centers.

Syringe Services Program: Addresses of all programs, provide access to sterile needles and syringes free of cost and facilitate safe disposal of used needles and syringes. Treatment Centers: Addresses of all opioid use disorder treatment centers. Acute Care Hospitals: Addresses of all acute care facilities.

Where are emerging opioid overdose hotspots in Massachusetts? The Massachusetts Department of Public Health compiles a database that com-bines public health records, including data from the state prescription monitor-ing program (PMP). Analyses for this poster used PMP data, to calculate Poten-tially Inappropriate Prescribing (PIP) variables, based on prior work by Rose et al.3, and overdose deaths by address, were provided by the Bureau of Vital Sta-tistics. See Variables section for definitions. There were 1,688 Opioid Overdose Fatalities in 2015; 2,025 in 2016; and 1,862 in 2017 (N= 5,575). This poster tests a 3-pronged approach to understanding the opioid crisis by examining results from: Emerging Hotspot analysis, Local Outlier Analysis and a Vulnerability score, results may be valuable but may not be conclusive.

Overall Vulnerability Scores by Municipality

Highest Vulnerability Lowest Vulnerability

7.47 Chatham 1.94 Brookline

6.75 Plymouth 1.94 Cambridge

6.62 Brimfield 2.00 Arlington

6.51 Webster 2.00 Belmont

6.50 Harwich 2.00 Somerville

Overall Vulnerability Results The all-variable vulnerability map shows trends of high vulnerability in Plymouth, eastern Cape Cod and Nantucket. Central Massachu-setts, west of Worchester and east of Springfield was another area with overall risk and lastly southwestern Massachusetts also had some high pockets of overall vulnerability. It is important to note that while Boston shows trends in overall low-vulnerability, that is probably more due to high resources, rather than low risk, because we know from the Hot Spot analysis that there are already high counts of fatalities in Boston and its surroundings.

Vulnerability to PIP Results By isolating just high rates of fatalities and high PIP, different areas emerge as vulnerable. Notably, Springfield, North Hampton, Am-herst, Deerfield and Greenfield, appear to have higher vulnerability to PIP and fatalities. Plymouth and most of Cape Cod also have high PIP vulnerability.

Vulnerability to Resources Results The resource vulnerability map shows where there are both high rates of fatalities and low resources or longer distances to resources. The low resource map shows patterns of vulnerability in southwest-ern Massachusetts, central Massachusetts, east of Amherst and West of Worcester.

PIP Vulnerability Scores by Municipality

Highest Vulnerability Lowest Vulnerability

8.01 Plymouth 1.06 Orleans

7.92 Chatham 1.15 Eastham

7.79 Yarmouth 2.09 Belmont

7.73 Dennis 2.27 Wellesley

7.49 Goshen 2.34 Wellfleet

Resource Vulnerability Scores by Municipality

Highest Vulnerability Lowest Vulnerability

7.55 Sandisfield 1.22 Belmont

7.41 Mount Washington 1.26 Watertown

7.23 Tolland 1.27 Tisbury

7.2 Otis 1.28 Arlington

6.69 Nantucket 1.41 Newton

Discussion Limitations: The PIP variables were only available at ZCTA level spatial unit, with fatality data and resources were address level. A more precise spatial location of PIP practices would provide a more precise vulnerability score and analysis, however these were the best available data. Strengths: This overall threefold analysis, using counts of fatalities for hotspot and local outlier analysis and rates of fatalities for vulnerability analysis allows readers to understand the epidemic from both perspectives of counts and rates.

Conclusions

The Hotspot and Local Outlier analyses using counts, shows that the Boston area is experiencing high counts of mortali-

ties, whereas the vulnerability analysis, which uses rates, identifies other parts of the state as vulnerable. This experi-

mental combination of results provides a nuanced picture of the problem, although results may not be conclusive.

Further investigation into high risk areas can be guided by the vulnerability scores. Additional burden and resource varia-

bles, as well as statistical analysis of variable interactions would aid further research.

PIP Vulnerability Score by Municipality High PIP and High Fatality Rates

Plymouth

Boston Boston

Plymouth

Results

Variables

Methods for Vulnerability Analysis