Determining the Impact of the Opioid Crisis on a Tertiary-Care Hospital in Central New York to Identify Critical Areas of Intervention in the Local Community

Background Central New York has been afflicted by the heroin epidemic with an increase in overdose deaths involving opioids. Objective The objective of the study was to understand the epidemiology of hospitalizations related to a diagnosis of opioid use (OU). Design The study was designed as a retrospective analysis of hospitalized patients admitted from January 1, 2008, to December 30, 2018, using ICD-9 and 10 codes for heroin or opiate use, overdose, or poisoning. Setting. The study was conducted in a tertiary-care and teaching hospital located in Central New York. Patients. Hospitalized patients were included as study participants. Results Opioid use-related admissions increased from .05/100 hospital admissions in 2008 to a peak of 2.9/100 in 2018, a 58-fold increase. There were 49 deaths over the 11-year period for an overall case fatality of 1.2 per 100 OU admissions. The median age for all years was 40 years (SD of 13.7 years), and admissions were largely white caucasians (67.0% of all admissions). The mean length of stay was 8.55 days (SD 12 days), with a range of 1 to 153 days. The most frequent discharge diagnosis was due to infections (15.0% of discharge diagnoses) followed by trauma (5.8% of discharge diagnoses). Methicillin-resistant Staphylococcus aureus was more common in patients with OU (58.1%) than in patients with non-OU (43%) (p < 0.0001 by chi-square with Yates' correction). Spatial analysis was performed by zip code and demonstrated regional hotspots for OU-related admissions. Limitations. The limitations of this study are its retrospective nature and largely numerator-based analysis. The use of ICD codes underrepresents the true burden due to underreporting and failure to code appropriately. This study focuses on patients who are hospitalized for a medical reason with a secondary diagnosis of opioid use and does not include patients who present to the emergency room with an overdose underrepresenting the true burden of the problem. Conclusions Our results demonstrate the impact of the opioid epidemic in one tertiary-care center and the need to prepare for the costs and resources to address addiction care for this population.


Introduction
e United States (US) and Central New York (NY) are in the midst of an unprecedented epidemic of substance use disorder (SUD) specifically to heroin and other opioids. During 1999 to 2015, approximately 568,699 people died from drug overdose in the United States [1,2]. From 2014 to 2015, there was an increase in overdose deaths of 11.4% with 52,404 deaths in 2015 [1]. e introduction of fentanyl to heroin mixtures, in particular, resulted in a deadly combination and contributed to the increasing death rate [3]. Central NY has also been afflicted by the heroin epidemic with a dramatic increase in overdose deaths involving opioids from 6/100,000 in 2010 to 14/100,000 in 2015 [4]. e US opioid epidemic highlights the racial, age, and geographic disparity of adults with SUD and deaths from overdoses. For example, in young adults aged 20-34 years, the rate of overdose deaths was 14.2 per 100,000 and double the rate among older adults aged 55-64 years (6.3/100,000) as reported in the 2017 Annual Surveillance Report of Drug-Related Risks and Outcomes [5]. In New York State during the years 2013 to 2015, there were a total of 1,440 overdose deaths [6]. Analysis of mortality rates by regions in New York demonstrated disparities in deaths related to opioid use, with clusters of high mortality in the Hudson Valley, Catskills, western Central New York, Finger Lakes, ousand-Island Seaway regions, Richmond County, and Long Island [6]. Examining mortality rates as compared with prescription rates for opioids, the ousand Islands-Seaway region and Central NY, in particular, had the most number of counties with higher opioid-related mortality rates but lower opioid prescription rates, which may suggest that as prescription opioids dropped, alternatives such as heroin were being used.
Overdose deaths from opioids are the tip of the iceberg of total health-associated complications from opioids and specifically from intravenous (IV) drug use. Infections related to IV drug use (IVDU) include viral infections such as human immunodeficiency virus (HIV), hepatitis C virus (HCV), and hepatitis B virus (HBV), as well as bacterial infections of the heart valve (subacute bacterial endocarditis), spinal abscess, and osteomyelitis, necrotizing fasciitis, myositis, and skin abscesses [7][8][9][10]. Many of the bacteria involved are common skin organisms including staphylococci and streptococci, although Gram-negative bacteria and fungi have also been implicated. Of particular concern is the emergence of methicillin-resistant Staphylococcus aureus (MRSA) as a cause of bacterial infection in drug users [11]. During 2005-2016, surveillance data from the Center for Disease Control's (CDC's) Emerging Infections Program (EIP) were analyzed and demonstrated that people who inject drugs were 16.3 times more likely to develop invasive MRSA infections than others [11]. Hospitalizations related to substance use disorder from opioids and complication of IV drug use are not reportable; thus, the true impact at the hospital level due to the opioid epidemic is not known. One study using a representative sample of US hospitalizations from opioid use and infectious complications found a significant increase in hospitalizations from 2002 to 2012 (301,707 to 520,275, respectively), with infectious complications also increasing (3,421 to 6,535, respectively) [12]. e economic cost associated with this was estimated based on the Consumer Price Index by the Bureau of Labor and Statistics and tripled, from $4,574,263,003 in 2002 to $14,850,435,892 in 2012.
Understanding the spatial epidemiology of substance use disorder either from emergency room visits or hospitalizations can be a powerful tool to identify regional hot spots to direct limited resources and educational targeting. In a study of prescriptions and overdose cases in Massachusetts by zip code, regional overdose clusters were identified in specific hot spot counties, providing an opportunity at the local level to influence public health decisions and execution of resources [13]. e number of hospitalizations from overdose and infections related to IVDU at our teaching hospital and Central NY is unknown and is the primary purpose of this study. It is imperative that we understand the epidemiology of opiate-related hospitalizations and infections to better prepare for services and empiric therapy for this population, to identify regional hot spots that can be a point source to target education and intervention, and to understand the growing rate of antibiotic resistance to implement hospital-specific infection control interventions.

Study Design.
is was a retrospective analysis, database review of hospitalized patients admitted to university hospital from January 1, 2008, to December 30, 2018, using ICD-9 and 10 codes for any hospital admission or discharge diagnosis for heroin or opiate use, heroin or opiate overdose, poisoning by opium, poisoning by heroin, poisoning by methadone, and opioid abuse (see supplemental Table 1 for ICD-9 and ICD-10 codes used).
is study focused on patients who were hospitalized for a medical reason with a secondary diagnosis of opioid use and did not include patients who presented to the emergency room with an overdose and not admitted for observation. e goals and analysis for this study were descriptive and to examine trends in our hospital of patients with any admission or discharge diagnosis related to substance use disorder. Additional analysis was performed to identify information that may be useful to our clinicians, administration, and the county public health department including changing demographics of this population, other medical issues including types of infections and MRSA frequency, trauma, discharge follow-up, and home address locations as an indicator of geographic areas for admissions. Records were identified using a database query, extracted, and de-identified. Microbial cultures and antibiotic resistance results were extracted from the Department of Clinical Pathology, Microbiology Section, and matched by patient medical record number to the database for the same time period. Case fatality was calculated based on in-hospital mortality of the study population.

University
Hospital. University hospital is located in Syracuse, NY, and the major teaching hospital of the State University of New York (SUNY) Upstate Medical University. is is a 735-bed facility and major referral hospital for the area for a higher level of care and includes a catchment area that extends north from the Canadian border, south to the Pennsylvanian border, west to Rochester, and east to Albany.

IRB Approval.
is protocol was reviewed and approved with waiver of requirement to obtain informed consent by the Upstate Medical University Institutional Review Board, #1355962-2.

Statistical
Methods. Data were analyzed for demographics and temporal and spatial trends with individual parameters examined by chi-square and analysis of variance (ANOVA), as appropriate, using GraphPad Prism version 5.02, GraphPad Software Inc., San Diego, CA, IBM SPSS version 25, Armonk, New York, and R version 3.3.3 published on 2017-03-06. In particular, R packages "zipcode" and "maps" were used for the zip code plot. e statistical analysis of the study was exploratory in nature, and the selection of the analysis results for reporting was subject to the authors' clinical expertise. e p values from the study were provided to indicate the probability of the observed data at random, without adjusting the multiplicity issue.

Number of Hospital Admissions per Year.
e number of hospital admissions with an OU-related admission or discharge diagnosis increased substantially from 2008 to 2018 (Figure 1 To account for changes in total hospital admissions that may bias these results (total hospital admissions have remained constant through the study year), total OU admissions were calculated as a proportion per 100 total hospital admissions per year (Figure 1(b)). In 2008, this proportion was 0.05/100, which increased to a peak of 2.9/ 100 in 2018, a 58-fold increase in admissions due to OU. . Patient disposition on discharge included 415 (13.3%) who left against medical advice; 407 (13%) were discharged to a rehabilitation facility, psychiatric facility, or skilled nursing care facility; and 1,819 (58.4%) were discharged to home or self-care.

Demographic.
e gender differences between male and females who were admitted with a diagnosis of OU fluctuated from 2008 to 2018. e proportion females in 2008,2009,2010,2011,2012,2013,2014,2015,2016,2017, and 2018 was 50%, 30%, 28%, 45%, 33%, 38%, 51%, 38%, 46%, 43%, and 47%, respectively. For all years, the proportion of females and males was 44% and 56%, respectively. When examining gender differences during 2008 to 2013 as compared with the peak years in 2014 to 2018, the proportion of females increased 8.7% (p � 0.011 from Pearson chi-square test) from 36.1% during 2008-2013 to 44.8% during 2014-2018. e median age for all years was 40 years (SD of 13.7 years). is did not significantly differ by year (data not shown). Examining age and gender differences between the low-admission years of 2008 to 2013 to the high-admission years of 2014-2018, there were significant differences in age by gender. As shown in the box plot of age by years (Figure 3), yearly median ages are very similar within each time period. Median age drops 7 years (p � 0.001 from Wilcoxon rank-sum test), from 47 (IQR 32-54) during 2008-2013 to 40 (IQR 31-51) during 2014-2018. Figure 4 demonstrates that for females, the median age drops 11 years, from 51 (IQR 36-58.5) to 40 (IQR 31-53) years. For males, the median age drops 5 years, from 45 (IQR 31-54) to 40 (IQR 31-50) years. Two-way ANOVA on log-transformed age confirms that patients become younger after 2014 (p � 0.009), and the age difference is mainly contributed by females (p � 0.015). Admissions for OU by race primarily were in the white caucasian population, comprising 67.0% of all admissions followed by African Americans (17.6%), American Indians (1.3%), and others in the remainder of race categories. is did not vary by year (data not shown).

Length of Hospitalization.
e length of hospitalization from time of admittance to discharge was calculated for each year. Overall, the mean length of stay was 8.55 days (SD 12 days), with a range of 1 to 153 days. e overall mean length of stay by year did not differ statistically with a range of 6 to 9 days (data not shown). When looking at the length of stay for the low-admission years of 2008-2013 to the high-admission years of 2014 to 2018, there was a significant increase in length of stay from a mean of 6 (SD 0.1) days in 2008-2013 as compared with 8.8 (SD 12.2) days in 2014-2018 (p � 0.001, T-test).

Spatial Analysis. Spatial analysis was performed by zip code and demonstrated in
For all years, the largest number of admissions occurred in zip code 13204, Solvay region, Syracuse, with 299 admissions (10% of total). is was followed by zip codes in a descending number of total admissions, 13202 (213, 6.8%), 13203 (185, 5.9%), 13208 (173, 5.6%), 13205 (167, 5.3%),   Figure 3 demonstrates the spatial distribution for all years in Central New York, largely representing the referral area of university hospital with a concentration in and around the city of Syracuse.

Discussion
Our results reflect the experience of one tertiary-care center in Central New York during the national opioid epidemic to identify local areas of need and interventions to this public health problem. Our findings demonstrated that we have experienced a dramatic increase in the number of hospital admissions of patients with opioid use (OU), specifically related to heroin and intravenous drug use (IVDU). Our experience is largely reflective of the national experience with a large proportion of OU patients admitted with complications of infections related to IVDU [12]. Our findings are surprising with respect to the large number of trauma-related patients with a diagnosis of OU. We demonstrated that the proportion of MRSA isolates in patients with OU are significantly higher than the general hospitalized population reflective of the national experience on the emergence of risk of MRSA in drug users [11]. Spatial analysis of hospital admissions suggested a high concentration of admissions from specific locations within our area, providing an opportunity to target these local communities with education and interventions.
To address this epidemic requires close coordination between local, state, and federal programs focused on identifying individuals with substance use disorder and providing them with the resources to address their addiction. e focus, and appropriately so, has been on opioidrelated deaths and funding for Narcan training of first responders and local providers. Hospitalized patients represent a large burden of complications of drug use and an unrealized opportunity to provide resources to address their addiction while hospitalized. In our hospitalized data, the large majority of patients were discharged to self-care. is may be from patient choice, but from our experience, more should be offered to patients to address their addiction. Such efforts are frequently limited because of a lack of funding and lack of emphasis from local and state governments. ere should be close coordination with the local public health department to report hospitalizations and to integrate patients into community resources to address their addiction. Spatial analysis of hospital admissions may provide evidence of local communities that could be targeted for education and interventions, especially when resources are limited.
Our analysis identifies areas of improvement in the care of patients with OU and the need for the education of healthcare providers in recognizing and treating addiction. Basic demographic analysis of this population provides important information on the changing demographics of those with OU and valuable information for providers to understand to provide early recognition and treatment for addiction and for administrators to improve resources and addiction services. Addiction services for hospitalized patients, training for medical residents, and addressing barriers for treatment post-acute care have clearly been shown to result in recovery and sobriety for patients suffering from OU [14][15][16][17].
e limitations of this study are its retrospective nature and largely numerator-based analyses. e use of proportion of hospital admissions has an inherent bias and may not reflect the true changes in the population, as a populationbased rate would. ere is an inherent bias during admissions based on the patient location and referrals. For example, one patient may have had multiple hospital Journal of Addiction admissions that in this analysis considered as separate admissions. Another major limitation is the use of ICD codes to identify patients, as this largely underrepresents the true burden because physicians underreport and often fail to code for opioid use. e change in ICD codes from ICD-9 to ICD-10 may also result in potential bias and an increase in diagnosis of substance use-related diagnoses. Spatial analysis though useful in this analysis has inherent bias, not controlled for in this study by using population-based rates.
We believe an analysis such as this should be considered by every hospital where opioid deaths are occurring. e information gained is valuable for hospital administrators and physicians to plan for the costs and resources needed for the care of this population and to be proactive in addressing this national health crisis.

Data Availability
Data availability is limited due to IRB rules and protected information.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this paper.