Pittsburgh's Opioid Epidemic

In 2017 Allegheny county saw a record 737 deaths from drug overdose, of those 623 coming from opioid overdose. While these numbers are shocking they are not unprecedented.

In 2014, following the rapid expansion of prescriptions for opioids, synthetic opioids like fentanyl and carfentanil entered the United States. Since their arrival in 2014 the number of fatal annual drug overdoses nationwide have doubled. 

Pittsburgh has labored to battle this epidemic. Using criminal and public health data from 911, Emergency Medical service, and Allegheny county coroners we will explore the effects of the Opioid epidemic on Pittsburgh and greater Allegheny county

In 2018 drug overdoses fell to 423 total deaths of those 366 coming from opioid overdose. Where were overdose calls coming into to EMS, and where are fatal overdoses occurring?

The Western Pennsylvania Regional Data Center collects and publishes all calls to Emergency Medical Services. We can use these calls as a way to better understand where overdoses are occurring and how frequent they are relative to the population.

When we normalize for population we see the majority of EMS calls for overdose are occurring in the city of Pittsburgh and the surrounding suburbs of McKees Rocks, and Braddock, with the epicenter being in downtown Pittsburgh.

While EMS calls provide some insight into the locations of the drug epidemic in Allegheny county, we still are not able to easily identify the areas where overdoses are more likely to be fatal and which communities are at highest risk.

To better understand where overdoses are most likely to be fatal we need to look at coroner data. Where did the 423 overdoses that occurred in 2018 happen? EMS data shows high numbers of overdose calls to EMS in downtown Pittsburgh, Braddock, McKees Rock, East McKeesport and throughout the south of the city.

Adding fatal overdoses to the map shows the highest counts of fatal overdoses within Pittsburgh city limits, and the suburbs of McKeesport and McKees Rock, however there is evidence of high overdoses west of the city in Coraopolis. In order to have greater granularity areas are shown in zipcodes rather than neighborhoods. However since the bulk of the overdoses are in Pittsburgh city limits lets focus on Pittsburgh.

Zooming into the neighborhood level we see that EMS calls in 2018 were centralized in the Central Business District, Crawford Roberts, the Strip District . Middle Hill and Central Oakland.

However when we scan to the map of fatal overdoses we find the highest counts in East Hills, Glen Hazel, Hazelwood, Marshall Shadeland, the Northside, East Liberty and Highland Park. We also see high rates in southern Pittsburgh like Beltzhoover, Mount Washington, and Carrick. Fatal overdoses rates downtown and in the Strip District are actually low relative to the surrounding neighborhoods.

One response to the epidemic has been an increased focus on arrests for drug possession and intent to sell, but it is unclear if arrests have been effective in slowing the growth of opioid epidemic. Each yellow diamond represents an arrest that included a 13(a)(16) possession of controlled substance charge. We see the largest clusterings in Homewood and the Northside. Whereas the Northside suffers from high counts of overdoses, so does Beechview however the number of arrests is much lower. Homewood and the Northside are demographically largely black neighborhoods, and it suggests there may be bias in policing affecting the pattern of arrests we're seeing.

By drilling down to the block level we can see how effective preemptive policing measures have been in curbing Pittsburgh's drug epidemic and if there is any potential bias revealed by the differential between overdose calls to Emergency Medical Services and arrests made for drug possession of a controlled substance.

The map reveals a troubling trend. Arrests being made for possession of a controlled substance are being made at higher frequency in primarily black neighborhoods. The overlaying of EMS overdose call data shows not only are these arrest rates disproportionate to number of overdoses in the area, but that in whiter neighborhoods that have higher quantities of overdose arrests are not often made. We see areas like Hays, New Homestead, and Mount Washington with few arrests but rather high overdose rates. This is also true of the areas surrounding the colleges like Squirrel Hill and South and Central Oakland.

In 2019 Allegheny county fared worse in battling the opioid epidemic than in the year prior. 564 people died of drug overdose in Allegheny county that year. Pittsburgh is the epicenter of the county and understanding trends that exist within it may be able to help us understand the course of the epidemic.

Some clear trends emerged from observing 2018, but to see if these are persistent trends and not simply outliers we can compare the data from 2019 to 2018. Reviewing the data from 2019 we see that neighborhoods with high frequencies of overdose calls expanded out from the the downtown area and central Oakland as we saw in 2018 into the the Northside in neighborhoods like North Shore, East Allegheny, and Marhsall-Shadeland, and expanded as well into the Southside. Additionally we see the frequency of overdose calls increasing in Homewood North and Point Breeze suggesting the Point Breeze area may be an emerging hotspot of overdoses.

We see the arrests patterns match the newly emerging overdose patterns in Northside, and a clustering of arrests in the Southside as well. We also can see a shift in both overdoses and arrests that began to become apparent in 2018, from the East Hills to Point Breeze. In 2018 while arrests were high in the East Hills and fatal overdoses were high, overall overdose calls relative to the population were low. The opposite however was true in Point Breeze in 2018 with high overdose calls and low arrests. A year later in 2019 we see arrests for drug possession shifting from the East Hills into the Point Breeze area, along with a similar shift in overdose calls to EMS. This suggests a change in the area of sale of drugs either in response to the increased police presence or in response to the high number of fatal overdoses in the neighborhood.

Reviewing fatal overdoses we see an epidemic that runs along the spine of the city, while areas like the Northside, Mount Washington, and Beechview had a lower amount of fatal overdoses relative to other areas than they did in 2018 they remain areas of concern. The most dire circumstance seems to be occurring in the Carrick area where overdose calls to EMS are high, and fatal overdoses continually top those of other areas of the city.

We see areas on the edges of the city have high frequencies of fatal overdoses as well, with areas like New Homestead, East Hills, East Carnegie, and Summer Hill all having relatively high totals of fatal overdose. Looking at the map above we see this trend is consistent when considering the number of calls to EMS for overdoses as well.

Showing the entirety of Allegheny county adds additional context showing nearly all neighborhoods bordering the city have high fatal overdose counts. This suggests a fundamental problem with the manner in which policing around drugs is conducted. Pittsburgh police share open data with the public of the arrests they make, however a significant amount of the demand for illicit drugs, namely opioids appears to be on the outskirts of the city. The opioid crisis appears to be emanating out from the edges of the city to supply a base of users in the surrounding suburbs.

Police data for those neighborhoods is not shared publicly, but to gain a greater understanding of the patterns of arrests made by police we can use multivariate clustering analysis to group arrests. We can see that use is high in the surrounding suburbs, but are illicit drugs flowing in from the outside or out from the inside?

The pattern of arrests tells us police are primarily focused on the Homewood and Northside communities. However EMS overdose calls are high throughout Marshall Shadeland, Perry North, Fairywood, Oakwood, East Carnegie, Carrick and North and South Oakland. Fatal overdoses are highest in Carrick, Brookline, Beechview, Mount Washington, and New Homestead, as well as the Northside.

These neighborhoods happen to be majority black neighborhoods, suggesting a pattern of racial bias may be affecting arrest rates in Pittsburgh. To determine this I introduced a k cluster with k=8 groups. To indicate these groups I used the following schema:

Shape :

• Circle = White
• Diamond = Black
• Triangle = Asian or Other

Color:

• Red = Young (~23 - 31 years of age on average)
• Blue = Female
• Yellow = Older (~52 -55 years of age on average)
• Orange = Middle aged (black men only ~ 36 years of age on average)
• Green = Asian or Other

It is clear that the number of diamonds indicating the person arrested was black far exceeds the all other racial groups. This is further exacerbated by the demographic breakdown of the city of Pittsburgh. Estimates place the Pittsburgh population at 300,286 residents of which 23.0% are black. This corresponds to approximately 69,065 black residents in the city of Pittsburgh. Of the 1,591 arrests for drug related charges in Pittsburgh 939 were of black residents, whereas only 593 were of white residents. This equates to 1 in 73 black residents being arrested for drug charges in 2019 alone, compared to a rate of 1 in 338 of the 200,591 white residents.'

Data from EMS overdose rates shows overdoses occurring with high frequency in neighborhoods of similar economic backgrounds with majority black residents occurring at equal rates to those of white residents. Fatal overdose data shows 91 of the 495 fatal overdoses in Allegheny county were of black residents and only 63 of 289 or approximately 21.8% of overdoses in the city limits were of black residents. White residents accounted for 222 of the 289 or about 76.8% of fatal overdoses in 2019. Black fatal overdose rates are in proportion to their population and white overdose rates are slightly above their population expectation. Arrest rates however are the converse with black residents representing 59.0% of drug arrests, with white residents representing 37.3% of arrests.

The map shows a persistent pattern of bias against black residents in arrest rates in the Pittsburgh city limits that is not explained by overdose rates, suggesting systemic over-policing of black neighborhoods. This disparity primarily impacting black men between the ages of 20 and 40. This impact is felt in particular by the Homewood community and the community in the Northside.

On March 11, 2020 the World Health Organization declared the newly emerged SARS-COVID-19 virus a global pandemic. The spread of the COVID-19 pandemic across the globe caused the most significant shift in global priorities since World War II, and significantly disrupted daily life all over the world. Pittsburgh was not exception with children being kept home from school, office buildings closing, universities shutting down, and all activities being shuttered for a period of months. The institution of lockdowns led to periods of prolonged isolation and following the announcement of the global pandemic in March 2020 drug overdose rates spiked to record levels across the United States. In the first 6 months of the pandemic 48,873 Americans died of drug overdose with deaths reaching a record high of 9,301 Americans in May 2020. The majority of these deaths were caused by opioids both prescription and synthetic. Pittsburgh overdose deaths remained below their 2017 high but still rose precipitously. To understand the impact on the Pittsburgh community we can look to overdose data.

In the year 2020 there were 545 fatal overdoses in Allegheny county an increase over both 2018 and 2019, but not as stark an increase as the country on average. Allegheny county shares open data on annual overdoses for each year that includes age, racial demographics, gender demographics and toxicology data. In order to protect the identity of the deceased no names are shared and location is only filtered down to the zipcode. However zipcode information is specific enough to provide crucial insights into understanding the opioid epidemic. The opioid epidemic in Pittsburgh is one primarily born of synthetic opioids, specifically fentanyl. Fentanyl or one of its close chemical cousins such as carfentanil was responsible for 480 overdose deaths in the year 2020, about 88% of total overdoses. 450 of these overdoses occurred after the declaration of the pandemic on March 11, 2020, with deaths peaking in April the first month of lockdown at 58 within the county. While synthetic opioid use was nearly universal, the combination of other drugs taken with it was not.

To explore the types of users overdosing and the flow of narcotics and other drugs throughout the county I ran a k cluster on county wide fatal overdoses setting k = 6. From this cluster I was able to identify 6 types of drug users each corresponding to one of the colored diamonds.

Benzodiazepine dominant users - Blue Diamonds

Fatal overdoses in this category all used a benzodiazepine that contributed to their overdose, with the group wide average being 1.034. These users were almost entirely white, and about 62% of which were male. These users had the highest average synthetic opioid usage at ~1.138 per overdose, second only to the primarily opioid users. Benzodiazepine contributed overdoses had the highest average rate of prescription opioid use of all 6 groups and approximately 48% of users in this group concurrently used heroin. This group also had a slightly higher than average use of antidepressants than the other groups. However both cocaine and methamphetamine use was lower than that of other groups. Highest frequency of these types of overdoses was found in 15116 - Glenshaw, 15210 - Mount Oliver, 15136 - Mckees Rock, and 15237 Ross Township, which are clustered above the north of the city.

Fentanyl and Heroin dominant users (male) - Green Diamonds

Fatal overdoses in this category are by far the most common with 195 overdoses falling into this group. The average age of users in this group was 41.13 years of age nearly identical to that of the benzodiazepine dominant users, however all overdoses in this group are male, and approximately 75% are white. Synthetic opioid use was nearly ubiquitous with many overdoses in this category using more than one synthetic opioid at once. This group had the highest average synthetic opioid use with ~1.26 per overdose. Approximately 26% of users in this group used prescription opioids, however alcohol user was extremely low with only 2 of 195 overdoses being identified positively for alcohol. The group also reported the second lowest cocaine usage, second only to the benzodiazepine grouping. No users in this group used benzodiazepines,, but this group reported significantly higher heroin use than other groups with 56.4% of users using heroin. Methamphetamine use was also low at 10.7% of users. 2 Approximately 20% of users in this group used only synthetic opioids. While these users were the most common in the sample there are particularly large and concentrated clusters of this group of users in the south of Pittsburgh in the Carrick, Brookline, Beechview, and Banksville neighborhoods, as well as in 15205 - Crafton. This suggests the high number of overdoses in Pittsburgh come primarily from this group as these neighborhoods have some of the highest fatal overdose rates in the city. Based on the initial findings I suspect these users were predominantly heroin users who graduated to synthetic opioids, or are users who were not aware they were purchasing heroin cut with synthetic opioids .

Alcohol and Cocaine dominant users - Purple Diamonds

Fatal overdoses in this category had the highest average age of 47.16 years. 84% of users in this group were male and this group had the highest diversity of racial background with only ~45% of users being white. This was the only group to have an average synthetic opioid use per overdose below 1 at 0.881. Overdoses in this group also had one of the lowest reported uses of prescription opioids with only between 3 and 5 overdoses having a contributing factor of prescription opioids. Alcohol use in this group was significantly higher in this group than other other groups with 79.57% of overdoses involving alcohol. Cocaine use was also higher in this group with 62.37% of overdoses involving cocaine, a significantly higher proportion than all other groups. This group reported no benzodiazepine usage and one of the rates of heroin usage at 30.11% Use of all other drugs was low. Additionally 14 of the 93 users in this group were cocaine only users and did not use opioids of any kind. The largest clusters of this group appear in Homewood, the Southside, Highland Park, and East Liberty. Outside of city limits clusters of this group appear in Mckees Rock and Monroeville.

Fentanyl dominant users (female) - Yellow Diamonds

This cluster is almost entirely female, and is the third largest of the three with 89 overdoses classified into this group. Users in this group had the lowest average age of all groups at 39.84 years of age. Approximately ~79% of users in this group were white a similar proportion to the male dominant fentanyl users. Synthetic opioid use averaged just above 1 per overdose at 1.03 per overdose. Prescription opioid use mirrored that of male dominant fentanyl users at 26.97%. Alcohol use was similarly low at 14.60% with slightly higher cocaine contribution to overdose at with 43.82% testing positive for cocaine. Heroin use was significantly lower than male users at 39.33%, and nearly identical rates of methamphetamine usage. This was also the only other group with synthetic opioid only users which made up 16.85% of users.

Methadone users - Pink Diamonds

The smallest of the groupings was the methadone user group , with only 20 overdoses in this group, 19 of which were white. This group was evenly split between women and men and all but one user in this group were synthetic opioid users. Prescription opioid use in this group was low with only 2 overdoses in this group showing presence of prescription opioids. Benzodiazepine use however was common appearing in 9 of the 20 in this group. Half of the users in this group showed heroin use and 5 of the 20 had presence of anti-depressants, and 8 of the 20 had cocaine in their system at the time of overdose. Methadone users are found in the eastern half of Allegheny county with many appear in the surrounding suburbs of Highland Park and Lawrenceville, but generally dispersed in the eastern and southern outskirts of Pittsburgh city limits. These users were likely in the process of attempting to get clean and either faced relapse or accidentally overdosed.

Methamphetamine users/ Other drugs - Light Blue Diamonds

Fatal overdoses in this category were approximately evenly split among genders with ~46.15% being female. Overdoses in this group reported average synthetic opioid use per overdose of 1.077,, which is roughly equal to that of other groups classified. Prescription opioid use rates were similar to other groups as well with 23.08% showing prescription opioid use. Alcohol use was higher than other groups but nowhere near as high as the alcohol dominant group with ~23.08% of users showing alcohol use. Cocaine use was the lowest of all groups classified at ~35.90%. Heroin use was the lowest among this group of all groups classified with only ~28.21%. This group had significantly higher methamphetamine usage with 23.07% using. This group also had some of the highest rates of antidepressant usage with ~33.33% showing antidepressants in their toxicology reports. The most significant difference present among overdoses in this group was the presence of other drugs ranging from barbiturates to MDMA, with ~1.03 drugs that did not fit into any of the categories previously mentioned on average. The high presence of methamphetamine with antidepressants suggests users may have been attempting either consciously or subconsciously to self medicate. There is well known comorbidity of ADHD and depression which appears at higher rates in methamphetamine users and appears to be corroborated by toxicology of overdoses in this group.

When results are clustered it provides some context to the epidemic of overdose deaths in Allegheny county which rather than battling a single drug epidemic seems to be experiencing numerous small epidemics that can centralized to specific regions. Carrick, Beechview, and Brookline have a high proportion of heroin users overdosing from synthetic opioids indicating that heroin and synthetic opioids in the area are interchangeably sold. There is also a possibility users in these areas are not aware they are ingesting synthetic opioids at the time of use. Fentanyl due to it's low cost and abundance is frequently cut into heroin to increase the heroin's potency. Clustering indicates drug checking services would be of high value in the these neighborhoods as well as in suburbs surrounding the south and west of the city like Crafton and Mckeesport.

To investigate this theory I created an additional k-cluster exclusively on fatal overdoses within Pittsburgh city limits. I again used k =6 groups to identify user types within Pittsburgh city limits. I overlaid these clusters onto a choropleth map of EMS calls of overdose.

The 6 groups identified broke down similarly to those of the cluster above but with some greater specificity.

Meth Users: Blue Circles

Only 9 overdoses in the city of Pittsburgh in 2020 involved methamphetamine, suggesting methamphetamine use is not widespread in Pittsburgh, this was also true of greater Allegheny county. Of those who methamphetamine contributed to their overdose 2/3 were men and 7 of the 9 were white. All of the overdoses in this group concurrently were using synthetic opioids and 6 of the 9 were using cocaine. The majority of these overdoses were found in and around Southside, Beechview, and Mount Washington, with very little methamphetamine contribution outside of this area. When reviewing the clustering of Allegheny county we see methamphetamine use is similarly restricted to the south of the county, clustering in the areas mentioned as well as 15025-Clariton and 15132 - Mckeesport. These maps suggest a methamphetamine problem that is limited in scope to primarily the south of the county in areas that are majority white and economically disadvantaged. This may be an area that would respond well to targeted intervention.

Synthetic Opioid Only Users: Red Circles

In the above Allegheny county wide clustering primarily heroin users and primarily synthetic opioid users were grouped together. When filtering down to the city of Pittsburgh we see that synthetic opioid only users are separated from users of heroin. There were 11 overdoses that fit into this group and of the 11 classified 9 of them are centralized to just 4 neighborhoods Brighton Heights, Perry North, Sheraden, and Highland Park. The most significant clustering exists in Brighton Heights and the high density of the this group suggests the overdoses with purely synthetic opioid may have originated from the same source. The small geographic area of the synthetic only group may be indicative that the Brighton Heights area has sales of other more powerful opioids. Look directly at the data it showed these clustered correlated with presence of carfentanil. When comparing these clusters to the data from the fatal overdoses file we can see clusters of overdoses sharing carefentanil centralized to small geographic areas.

Multiple Addiction Long Time Users: Green Circles

Overdoses in this group are of users who are on average significantly older than other groups with an average age of 54.82 years, they're ~68% male and majority black. Average synthetic opioid use is approximately equal to the average of other groups a ~1.11 per overdose. This group has high cocaine use with ~80% showing cocaine contribution to their overdose. More than half also showed alcohol contribution to overdose. When comparing the average age of cocaine users it revealed a significant difference by by racial background with the average age of overdoses showing cocaine contribution being 8 and a half years higher at 48.5 years than those of white overdoses at ~40 years of age. Cocaine use was much more common in overdoses of black residents than white residents with 60.33% of overdoses of black residents showing cocaine contribution compared to 38.14% of white residents. The data suggests residents with existing addictions to cocaine may be being introduced to synthetic opioids as synthetic opioids enter the community.

Cocaine Only Users: Yellow Circles

Only 5 overdoses in Pittsburgh came from only cocaine use. This result is unsurprising. The spread of synthetic opioids across the country has taken over the market of drugs in the United States and Pittsburgh is no exception. As was mentioned above many of the existing users of drugs like cocaine and methamphetamine seem to be being pulled into the opioid epidemic as the wide availability of those drugs has made them ubiquitous.

Methadone Users: Purple Circles

Again methadone users were separated from other users in the process of clustering. Of the 8 overdoses that involved methadone all were of white residents. Synthetic opioid use was lowest among this group of all groups that used synthetic opioids, however benzodiazepine use was highest with 6 of the 8 using benzodiazepines. 1 of every 3 overdoses involving benzodiazepines also involved methadone. The lower rate of synthetic opioid use suggest a base of users trying to stop using synthetic opioids, and that concurrent users of benzodiazepine and synthetic opioids may be more likely to seek treatment. There is a troubling racial trend here as well with all methadone overdoses in Pittsburgh being of white residents and only 1 in all of Allegheny county being of a black resident. This suggests methadone clinics are not adequately reaching black residents and efforts to curb addiction to opioids in the black community must be stepped up by public health officials. Patterns of bias that exist in policing may also exist in substance abuse efforts as well.

Heroin to Fentanyl Users: Brown Circles

This group was the largest of all the groups of overdoses classified in Pittsburgh with 59 overdoses fitting this classification. The average age of this group was one of the youngest at 36.95 years of age, with only the synthetic opioid only group having a younger average age. Overdoses in this category were ~78% male and ~85% white. Users in this category also had the highest average synthetic opioid user per overdose of all groups with an average of ~1.29 per overdose, and all 59 overdoses using at least 1 synthetic opioid. this group also had the highest average prescription opioid use with ~33.90% using prescription opioids as well. However cocaine use was lower than other groups with ~45.75% of overdoses showing cocaine contribution. Heroin use was significantly higher than other groups with ~86.44% of overdoses showing use of heroin. While this group was omnipresent throughout the city we see the largest concentrations of overdoses in this group in the Beechview and Banksville area, the Brighton Heights area, and the Highland Park, East Liberty area. These results are consistent with the measurement of fatal overdoses throughout 2018, 2019, and 2020.

The drug epidemic in Pittsburgh is a complex and challenging issue without any quick or easy solutions. However there are steps that can be taken to confront the tragedy of Pittsburgh's opioid epidemic. Pittsburgh's opioid epidemic is not a single narrative, but rather multiple narratives that are occurring concurrently. Drug users in Pittsburgh are not a monolith and understanding users as one block may be counterproductive to curbing the loss of life from this epidemic.

The spread of fentanyl and the movement of carfentanil have been significant predictors to fatal overdoses in the Pittsburgh area. Neighborhoods in the Beechview and Banksville area as well as Carrick and Brighton Heights are of particular concern. Patterns of drug use differ in the city with high use of heroin and benzodiazepines in the south of the city and high use of heroin in the north of the city. The results of this study show that synthetic opioid use is widespread. The concurrent use of heroin and synthetic opioids is an enormous problem that has contributed to incredible loss of life both in and outside of the city. The area of Mckees Rock appears to be a significant area of distribution of drugs on the west and the Highland Park and East Liberty area appears to be a significant port of entry of synthetic opioids on the east.

Differences emerge among racial groups in pattern of use with overdoses of white residents showing high amounts of concurrent use of synthetic opioids and heroin and overdoses of black residents showing high amounts of concurrent use of synthetic opioids and cocaine. These trends may be contributing to public health inequities in the distribution of methadone among black resident versus white residents, and more coverage of black residents by methadone clinics is required.

While overdoses from synthetic opioids effect the entire community of Pittsburgh there is clear pattern of arrests among young and middle aged black men, that is not supported by increased overdose numbers or higher drug use. The Homewood area specifically appears to suffer from chronic over-policing of drug possession charges, however this is true more broadly for black men in general in Pittsburgh. This over-policing is counterproductive and damages the trust between communities and law enforcement. This trust is more important than ever as synthetic opioids enter these communities and significant public health campaigns are launched to combat their effects.