Police drug crackdowns and hospitalisation rates for illicit-injection-related infections in New York City

Methods 

The analysis consisted of five stages: (1) we identified the location and initiation dates of all crackdowns implemented in NYC during the study period and modelled their intensity; (2) we described the sociodemographic composition of each NYC precinct using United States (US) Census data; (3) we calculated precinct-specific monthly rates of hospitalisations for illicit-injection-related abscesses, cellulitis, and endocarditis using New York State hospitalisation data and US Census data; (4) we tested the hypothesis that the crackdowns were followed by an increase in the precinct-specific hospitalisation rates of the infections of interest; and (5) we augmented our statistical modelling with sensitivity analyses that attempted to account for the incarceration of injectors who developed an illicit-injection-related abscess, cellulitis, or endocarditis infection while living in the community and who would have contributed cases to the case count had they not been incarcerated before they could seek treatment at one of the community-based hospital facilities that were the source of our infection data (as described below).

All analyses were run on SAS 8.2 software. The Harvard School of Public Health Human Subjects Committee and the New York State Statewide Planning and Research Cooperative System (SPARCS) Institutional Review Board approved study protocols.

Stage 1: identifying the crackdowns’ locations, initiation dates, and intensities 

Data regarding the locations and initiation dates of drug crackdowns occurring in NYC between 1995 and 1999 were acquired through NYPD's Narcotics Division and newspaper reports of mayoral press conferences. To determine whether the intensity of the crackdowns changed during their implementation, we investigated arrest patterns in crackdown precincts, before and after the crackdowns’ initiations, using a New York State Division of Criminal Justice Services (NYSDCJS) database. NYSDCJS contains information on the major charge in the arrest (i.e., if an individual is arrested for two or more charges, only the most serious is reported), the month and year in which the arrest was made, and the precinct reporting the arrest (New York Division of Criminal Justice Services, 2000). NYSDCJS data, combined with information on precinct population size, were used to calculate the arrest rate for each precinct each month before and during each crackdown's implementation; the methods used to calculate precinct population size are described in the next section.

Exploratory analyses of these data suggest that arrest rates rose markedly in the months following the crackdowns’ onset. The median monthly precinct-specific arrest rate for drug-related offences climbed 39%, from 201 to 279 arrests per 100,000 precinct residents, comparing the 12 months preceding and following the crackdown's onset. Similarly, the median monthly precinct-specific arrest rate for all offences increased by 17% over the same time period, from 559 per 100,000 residents to 652 per 100,000 residents. As is evident in Fig. 1, arrest rates surged in the first quarter of the crackdown and then were relatively sustained thereafter. Additional analyses indicate that the quarter prior to the crackdowns’ onset saw a gradual ramping up in arrest rates in some crackdown precincts, whereas arrest rates in other crackdown precincts declined during these months, a pattern that suggests that the crackdowns were preceded by some type of preparatory period. Accordingly, we identified four apparent crackdown stages:

1)Baseline: 4–15 months prior to crackdown onset;

2)Preparation: quarter prior to crackdown onset;

3)Initiation: month of crackdown onset;

4)Crackdown quarters: four quarters following crackdown onset.

View Large Image Download to PowerPoint

Fig. 1. Median precinct-specific drug-related arrest rates and total arrest rates in the four quarters before and after the initiation of eight police drug crackdowns in 27 New York City police precincts.

Results 

Analyses suggest that sociodemographic characteristics of the 27 crackdown precincts in the first month of the study period (January 1995) differed considerably from those of the 49 precincts that did not undergo a precinct-specific drug crackdown between 1995 and 1999. Crackdown precinct residents tended to be younger, more impoverished, and were less likely to be non-Hispanic white than residents of non-crackdown precincts (Table 1). During this baseline period, median hospitalisation rates for illicit-injection-related abscesses, cellulitis, and endocarditis within crackdown precincts were double or more than double those observed for non-crackdown precincts.

Table 1. Sociodemographic characteristicsa and hospitalisation rates for illicit-injection-related abscesses, cellulitis, and endocarditis for all New York City police precincts, crackdown precincts, and non-crackdown precincts during January 1995

	Characteristics	All precincts (N=76)	Crackdown precincts (N=27)	Non-crackdown precincts (N=49)
	Precinct population size (mean, standard deviation)	101,000 (48,000)	89,000 (40,000)	107,000 (51,000)
	Precinct age structure (mean, standard deviation)
	0–17 years	23.1% (7.8)	29.0% (5.6)	20.0% (6.9)
	18–64 years	64.7% (6.7)	61.5% (4.3)	66.7% (7.2)
	>65 years	12.2% (4.0)	9.5% (2.3)	13.3% (3.8)
	Percent of residents subsisting below the federal poverty line (mean, standard deviation)	21.8% (12.3)	33.0% (11.3)	15.6% (7.7)
	Percent non-Hispanic white residents (mean, standard deviation)	38.2% (29.1)	12.0% (17.3)	52.6% (23.7)
	Crude monthly hospitalisation rates for illicit-injection-related infections (per 100,000)
	Abscess/cellulitis
	Mean (standard deviation)	22.9 (39.6)	24.4 (13.8)	22.0 (48.2)
	Median (interquartile range)	13.4 (9.2–25.7)	20.8 (13.2–34.4)	11.0 (7.8–20.3)
	Endocarditis
	Mean (standard deviation)	2.4 (2.9)	3.6 (2.4)	1.7 (2.9)
	Median (interquartile range)	1.5 (0.0–4.2)	3.3 (2.0–5.5)	0.0 (0.0–2.0)

a Precinct sociodemographic profiles for January 1995 were interpolated using 1990 and 2000 US Census block and block group data for New York City assuming linear change over time.

Bivariate analyses suggest that abscess and cellulitis infections followed a seasonal pattern, spiking during summer months (Table 2); endocarditis infections did not follow such a trend. Bivariate analyses also provide evidence of an underlying secular increase in infection rates within crackdown precincts, with abscess and cellulitis hospitalisation rates in 1999 substantially higher than in 1995 and endocarditis hospitalisation rates in 1999 substantially higher than in 1996. Rates of abscess, cellulitis, and endocarditis hospitalisations were approximately 1.8 times higher in precincts in which 20% or more of the population lived at or below the poverty line than elsewhere; the percentage of precinct residents aged 18–64 was inversely associated with hospitalisation rates for all infections studied. The percent of non-Hispanic whites residing in a precinct was inversely associated with endocarditis hospitalisation rates; while hospitalisation rates for abscess and cellulitis infections showed a similar trend, the trend did not reach statistical significance. Abscess, cellulitis, and endocarditis hospitalisation rates varied geographically, with rates higher in the boroughs of Brooklyn and Manhattan than in Queens.

Table 2. Crude and adjusted regressions of precinct characteristics, season, year, and crackdown stages on precinct-specific hospitalisation rates for illicit-injection-related abscesses, cellulitis, and endocarditis, among NYC's 27 crackdown precincts, 1995–1999

	Covariates	Hospitalisations for abscesses/cellulitisa			Hospitalisations for endocarditisa
		Crude relative riskb (95% confidence interval)	Adjusted relative riskc (95% confidence interval)	Adjusted P-value	Crude relative risk (95% confidence interval)	Adjusted relative risk (95% confidence interval)	Adjusted P-value
	Season
	Winter	0.75 (0.70–0.81)	0.76 (0.71–0.82)	<0.0001	0.86 (0.71–1.04)	0.84 (0.69–1.02)	0.08
	Spring	0.87 (0.81–0.93)	0.87 (0.82–0.91)	<0.0001	0.96 (0.82–1.12)	0.94 (0.81–1.10)	0.43
	Fall	0.84 (0.79–0.89)	0.86 (0.82–0.93)	<0.0001	1.14 (0.98–1.33)	1.18 (1.00–1.39)	0.05
	Summer (referent)	1.00	1.00	–	1.00	1.00	–
	Percent residents living below poverty level
	≥20%	1.88 (1.54–2.29)	0.67 (0.41–1.09)	0.10	1.83 (1.51–2.23)	0.93 (0.50–1.74)	0.83
	<20% (referent)	1.00	1.00	–	1.00	1.00	–
	Percent residents non-Hispanic white	0.99 (0.98–1.00)	0.99 (0.98–1.00)	0.002	0.98 (0.98–0.99)	0.99 (0.98–1.00)	0.007
	Percent residents aged 18–64 years	0.94 (0.91–0.97)	0.96 (0.95–0.98)	0.0002	0.94 (0.92–0.97)	0.98 (0.95–1.01)	0.15
	Boroughd
	Bronx	1.64 (1.21–2.22)	2.20 (1.29–3.76)	0.004	1.45 (0.97–2.18)	1.42 (0.78–2.58)	0.25
	Brooklyn	2.46 (1.95–3.10)	2.82 (1.66–4.81)	0.0001	1.81 (1.50–2.18)	1.53 (0.82–2.84)	0.18
	Manhattan	1.53 (1.26–1.85)	2.49 (1.46–4.26)	0.0008	1.48 (1.19–1.84)	1.86 (0.93–3.72)	0.08
	Queens (referent)	1.00	1.00	–	1.00	1.00	–
	Year
	1995	0.63 (0.47–0.85)	0.57 (0.38–0.84)	0.005	0.86 (0.63–1.19)	0.62 (0.32–1.22)	0.17
	1996	0.59 (0.44–0.78)	0.60 (0.45–0.82)	0.001	0.63 (0.46–0.88)	0.50 (0.30–0.86)	0.01
	1997	0.55 (0.41–0.73)	0.60 (0.48–0.74)	<0.0001	0.62 (0.44–0.87)	0.60 (0.41–0.90)	0.01
	1998	0.74 (0.59–0.93)	0.74 (0.63–0.86)	0.0001	0.74 (0.56–0.98)	0.68 (0.50–0.93)	0.01
	1999 (referent)	1.00	1.00	–	1.00	1.00	–
	Crackdown stagese
	Baseline (referent)	1.00	1.00	–	1.00	1.00	–
	Preparation	0.96 (0.90–1.02)	0.94 (0.85–1.04)	0.25	1.11 (0.88–1.40)	1.21 (0.96–1.54)	0.11
	Initiation month	0.99 (0.89–1.09)	0.89 (0.78–1.00)	0.06	0.92 (0.68–1.26)	0.96 (0.66–1.41)	0.85
	First crackdown quarter	1.06 (0.98–1.15)	0.88 (0.77–1.00)	0.06	0.92 (0.73–1.16)	0.91 (0.66–1.25)	0.54
	Second crackdown quarter	1.09 (0.98–1.22)	0.86 (0.74–0.99)	0.04	0.89 (0.73–1.09)	0.81 (0.62–1.06)	0.13
	Third crackdown quarter	0.94 (0.87–1.02)	0.80 (0.67–0.94)	0.008	0.87 (0.66–1.15)	0.74 (0.53–1.05)	0.09
	Fourth crackdown quarter	0.79 (0.71–1.15)	0.72 (0.57–0.91)	0.006	0.62 (0.44–0.89)	0.57 (0.35–0.92)	0.02

a 48,986 abscess and cellulitis cases and 5452 endocarditis cases aggregated to the precinct level. b “Crude” analyses can also be thought of as bivariate analyses. c “Adjusted” analyses included all variables listed in the table as model covariates. d No crackdowns occurred in Staten Island during the study period. e Crackdown stage definitions: Baseline: 4–15 months prior to crackdown initiation; Preparation: 1–3 months prior to crackdown initiation; Initiation month: month of crackdown initiation; Crackdown quarters: quarters following each crackdown's initiation.

Multivariate analyses based on infections among only the non-institutionalised population indicated that declines occurred in illicit-injection-related abscess, cellulitis, and endocarditis hospitalisation rates following the crackdowns’ onset, especially during the fourth quarter. Sensitivity analysis results (Table 3), however, suggest that hospitalisation rates for illicit-injection-related abscesses and cellulitis either declined only in the fourth crackdown quarter or remained static after the crackdowns’ onset, depending on whether we assumed that five or ten percent of incarcerated injectors were infected at incarceration. Hospitalisation rates for endocarditis remained static when we assumed that one percent of recently-incarcerated injectors were infected when incarcerated and increased during the first crackdown quarter when we assumed that five percent of these individuals were infected. In these multivariate models, precinct age and racial/ethnic composition, season, year, and borough remained statistically significant predictors of abscess and cellulitis hospitalisation rates; year and the percent of precinct residents who were non-Hispanic white remained significant predictors of endocarditis hospitalisation rates.

Table 3. Multivariate regression of crackdown stages on precinct-specific hospitalisation rates for illicit-injection-related abscesses, cellulitis, and endocarditis among New York City's 27 crackdown precincts, 1995–1999: with and without adjustments for incarceration of individuals harbouring infectionsa

	Crackdown stagesb	Hospitalisation rates for abscesses/cellulitis			Hospitalisation rates for endocarditis
		No incarcerated individuals infected relative risk (95% confidence interval)	5% of incarcerated individuals infected relative risk (95% confidence interval)	10% of incarcerated individuals infected relative risk (95% confidence interval)	No incarcerated individuals infected relative risk (95% confidence interval)	1% of incarcerated individuals infected relative risk (95% confidence interval)	5% of incarcerated individuals infected relative risk (95% confidence interval)
	Baseline (referent)	1.00	1.00	1.00	1.00	1.00	1.00
	Preparation	0.94 (0.85–1.04)	0.95 (0.86–1.05)	0.96 (0.87–1.05)	1.21 (0.96–1.54)	1.21 (0.97–1.51)	1.19 (1.01–1.40)
	Initiation month	0.89 (0.78–1.00)	0.91 (0.81–1.03)	0.93 (0.83–1.05)	0.96 (0.66–1.41)	1.01 (0.72–1.42)	1.11 (0.87–1.42)
	First crackdown quarter	0.88 (0.77–1.00)	0.95 (0.83–1.08)	1.01 (0.89–1.14)	0.91 (0.66–1.25)	1.02 (0.77–1.36)	1.36 (1.10–1.69)*
	Second crackdown quarter	0.86 (0.74–0.99)*	0.92 (0.79–1.06)	0.97 (0.84–1.12)	0.81 (0.62–1.06)	0.92 (0.71–1.18)	1.22 (0.99–1.50)
	Third crackdown quarter	0.80 (0.67–0.94)*	0.85 (0.72–1.01)	0.91 (0.77–1.07)	0.74 (0.53–1.05)	0.84 (0.62–1.15)	1.13 (0.88–1.44)
	Fourth crackdown quarter	0.72 (0.57–0.91)*	0.78 (0.62–0.97)*	0.83 (0.67–1.03)	0.57 (0.35–0.92)*	0.68 (0.45–1.03)	0.99 (0.73–1.35)

a Adjusting for year, season, precinct borough, precinct racial/ethnic and age composition, and precinct poverty status. b Crackdown stage definitions: Baseline: 4–15 months prior to crackdown initiation; Preparation: 1–3 months prior to crackdown initiation; Initiation month: month of crackdown initiation; Crackdown quarters: quarters following each crackdown's initiation. * P<0.05.

Discussion 

Study results based on data for the non-institutionalised population in 27 crackdown precincts in NYC did not support our hypotheses that police drug crackdowns would be associated with increases in precinct-specific hospitalisation rates for illicit-injection-related abscesses, cellulitis, and endocarditis, and instead yielded evidence of a decline in these rates. Additional sensitivity analyses that took into account the likely infection rate among incarcerated individuals, however, suggested that hospitalisation rates remained largely static following the crackdowns’ initiation.

Our findings regarding the relationship of crackdown stage to hospitalisation rates are consonant with several possible scenarios. First, it is possible that the prevalence of injection drug use declined in the 27 crackdown precincts as a result of the crackdowns’ onset. The increase in drug-related arrests after each crackdown's initiation would have diminished the pool of injectors living in the precinct during the four quarters after each crackdown began. The magnitude of this reduction would have grown with each passing crackdown quarter as more injectors were arrested, repeat offenders were incarcerated for longer and longer periods, and injectors who did not use drugs while incarcerated continued to abstain from drugs after their release from the criminal justice system. This scenario is consistent with the increasing declines in hospitalisation rates seen in each subsequent crackdown quarter compared to baseline rates. While the sensitivity analyses allowed for the possibility that a proportion of recently-incarcerated individuals harboured the infections of interest, they did not account for the transfer of a substantial proportion of the at-risk population from the police precinct to the criminal justice system.

Additionally, research regarding drug treatment initiation suggests that entanglements with the criminal justice system are treatment motivators (Shutz, Rapiti, Vlahov, & Anthony, 1994; Weatherburn & Lind, 2001). While research by Wood and colleagues in Vancouver and Weatherburn and Lind in Sydney found no relationship between crackdowns and treatment admissions (Weatherburn & Lind, 1997; Wood et al., 2004), the latter investigation was hampered by a lack of variation in arrest rates and the former interviewed injectors only 3 months after the crackdown's onset. In contrast, the present study focused on crackdowns that generated a 39% increase in the median monthly drug-related arrest rate during the year following the eight crackdowns’ initiation and we followed precincts for 12 months after each crackdown's onset. Users’ resulting intensified involvement in the criminal justice system, or heightened concern about impending involvement, might have encouraged some to reduce the frequency with which they injected, switch to non-injection modes of administration, or abstain from drug use altogether.

A corollary scenario is also possible: while the crackdowns might have encouraged some injectors to stop or reduce their injection drug use, other individuals might have continued to inject drugs frequently and, in these heavily-policed circumstances, might have engaged in more unsafe injection practices during the crackdowns than prior to the crackdowns’ onset. Such a scenario would be consonant with the sensitivity analyses indicating a decline or stasis in hospitalisation rates for illicit-injection-related infections after the crackdowns began and also with all past research on fear of arrest and injection practices among active injectors (Aitken et al., 2002; Bluthenthal, Kral et al., 1999; Bluthenthal et al., 1997; Bluthenthal, Lorvick, et al., 1999; Bluthenthal & Watters, 1995; Bourgois, Lettiere, & Quesada, 1997; Cooper et al., in press; Grund, Heckathorn, & Anthony, 1995; Koester, 1994; Maher & Dixon, 1999; Zule, 1992). Given that our study contained no individual-level data with which to provide insight into individual drug use practices, however, we could not test this scenario's validity.

As Wood and colleagues, Curtis, and others have found, intensified drug-related policing activities can displace drug activity to new areas (Best, Strang, Beswick, & Gossop, 2001; Cooper et al., in press, Curtis, 2003, Wood et al., 2003, Wood et al., 2004). Three forms of displacement are possible, each of which may have exerted a different effect on rates of injection-related harms in crackdown precincts. First, injectors with access to private space may have shifted their drug use from parks and alleyways to private spaces within crackdown precincts to avoid police attention (Cooper et al., in press, Curtis, 2003). These injectors may have been better able to reduce the harm of their drug use in these private spaces than in public spaces, where access to running water and other materials needed for safe injection is limited or non-existent. Second, some crackdown residents may have started using drugs in non-crackdown precincts after the crackdown's onset, thus perhaps minimising their fear of arrest and consequent adverse health effects (Best et al., 2001, Curtis, 2003, Wood et al., 2003, Wood et al., 2004). It is also possible, however, that this displacement may have increased the likelihood of injecting outside (Wood et al., 2004), and thus increased the likelihood of unsafe injecting, as injectors began using drugs in unfamiliar neighbourhoods where they had few community ties. Finally, some injectors may have shifted not only the location in which they used drugs but also their residence from a crackdown to a non-crackdown precinct. Analyses of hospitalisation rates for illicit-injection-related infections in non-crackdown precincts adjacent to crackdown precincts, however, suggest that such displacement was insignificant.

Additionally, the estimated decline or stasis in hospitalisation rates after the crackdowns’ onset might partially be an artefact of the methods we used to identify cases of illicit-injection-related infections. Based on prior knowledge about typical sites of illicit-injection-related soft tissue infections (Ciccarone et al., 2001, Stone et al., 1990), our algorithm classified cases of abscess and cellulitis infection as illicit-injection-related if they occurred on the patient's extremities. It is possible, however, that drug-injecting residents might have endeavoured to avoid injecting in their extremities during the crackdown to minimise their visibility as drug users during a police search. A qualitative study exploring crackdowns and injection practices in one NYC crackdown precinct, conducted by the authors (Cooper et al., in press), suggests that some injectors employed this strategy to evade arrest. Similarly, we also used self-reported illicit drug use as a criterion for classifying infections as illicit-injection-related or not; it is possible that hospitalised individuals may have been more reluctant to disclose illicit drug use to health care personnel in the midst of the crackdown.

As stated earlier, we lacked access to HHC's inmate health data and thus used sensitivity analyses to investigate the effect on our analyses of excluding individuals who, had they not been incarcerated, would have remained eligible to be included in the SPARCS data for the non-institutionalised population. The elevated arrest rates found here following the crackdowns’ onset testify to the importance of taking the possible prison-based treatment of the infections of interest into account in the analyses, as does research suggesting that the quality of healthcare services provided within the criminal justice system may be similar or superior to that available in the community (Farley et al., 2000, White et al., 2001). Given the lack of actual data on infection rates among newly incarcerated individuals, however, our sensitivity analyses might have underestimated, or less plausibly, overestimated, the number of recently-incarcerated individuals harbouring the infections of interest.

Finally, unanticipated temporal trends in injection drug use, HIV/AIDS, and the incidence of hospitalisations for illicit-injection-related infections during the study period might have impeded our ability to detect crackdown-related changes in the outcomes. The number of injectors, and possibly those at particular risk of developing the outcomes, declined markedly citywide during the course of the study period. Research conducted by Friedman and colleagues and Holmberg suggests that the number of past-year injectors residing in the NYC metropolitan area declined by approximately 35% during the study period, from 197/10,000 residents in 1993 to 125/10,000 residents in 1998 (Friedman et al., 2004, Holmberg, 1993). Additionally, HIV incidence among injectors in NYC declined and highly active anti-retroviral therapy (HAART) emerged during the study period (Des Jarlais et al., 2000a, Des Jarlais et al., 2000b). Given that injectors with HIV, and perhaps in particular those with lower CD4 cell counts, are more susceptible to the infections studied than other injectors (Spijkerman, van Ameijden, Mientjes, Coutinho, & van den Hoek, 1997; Wilson, Thomas, Astemborski, Freedman, & Vlahov, 2002), these epidemiologic developments might have reduced the population at heightened risk of the study's outcomes during the study period. We were unable to incorporate these substantial reductions in the population at risk for illicit-injection-related infections in our statistical models; this limitation might have impaired our ability to compare trends in pre- and post-crackdown initiation hospitalisation rates, since we were analysing rates in relation to the total population in each precinct because no data exist on the size of the at-risk population (i.e., current injectors) in each precinct.

Similarly, the unanticipated citywide temporal increase in illicit-injection-related infections during the study period could have eclipsed any rise due solely to the crackdowns. This increase is of interest, particularly in a context of declining numbers of injection drug users in NYC. A contributing factor might be that injectors in NYC greatly increased their participation in syringe exchange programs during these years (Des Jarlais et al., 2000a, Des Jarlais et al., 2000b) and these programs may have linked injectors to hospital-based care for their infections. Research suggests that the rise in abscesses, cellulitis, and endocarditis found in NYC might not be a local phenomenon: a similar increase in hospitalisations for soft tissue infections occurred in San Francisco during the study period (Ciccarone et al., 2001). Similar to actions taken in San Francisco (Ciccarone et al., 2001), it may be useful in NYC for community clinicians, staff at programs serving injectors, epidemiologists, and active injectors to convene a task force to establish a research program to identify the causes of the increase in illicit-injection-related infections and formulate methods to address it. Given the morbidity and mortality associated with these infections, and the cost of treating them (charges for inpatient treatment of 945 individuals for soft tissue infections, the majority of which were injection-related, in one San Francisco hospital reached approximately (US) $9.9 million per fiscal year 1996–2000 (Ciccarone et al., 2001), such a task force should consider systematically monitoring patterns of hospitalisation for illicit-injection-related abscess, cellulitis, and endocarditis infections.

Study results suggest that further research into the effects of particular police strategies on injectors’ health is warranted. A useful line of inquiry lies in discerning which of the proposed scenarios described above, or which combination of these scenarios, is correct. Multi-level investigations examining the relationship between crackdowns and injectors’ health might be fruitful if they constructed the outcome as self-reported instances of illicit-injection-related infections and incorporated individual-level data regarding police encounters, injection drug use frequency and cessation, and injection practices and precinct-level data concerning the crackdown status of each participants’ home precinct and the precinct's sociodemographic profile. Additionally, as evidenced here by the elevated arrest rates following the crackdowns’ onset, future investigations should include information regarding inmate health status, acquired through processes that respect inmates’ human rights. Such research, coupled with past investigations into the relationship between arrest concerns and unsafe injection practices, could inform policy-makers entrusted with selecting drug-related law enforcement approaches about the possible unintended health consequences of the strategies under consideration.